BIO REGIONAL ISATION OF THE SOUTHERN OCEAN

REPORT OF EXPERTS WORKSHOP

(HOBART, SEPTEMBER 2006) a Wayne Papps, Australian GovernmentDivision, © Commonwealth of Australia Antarctic Wayne

© WWF-Australia and Antarctic Climate and Acknowledgements: Ecosystems Cooperative Research Centre (ACE CRC). All rights reserved. 2006. This report presents the outcomes of an Experts Workshop on Bioregionalisation of the Southern Ocean, held in Hobart, Australia, 4-8 September 2006, hosted by WWF-Australia and the Antarctic Climate and All material appearing in this publication is Ecosystems Cooperative Research Centre. copyrighted and may be reproduced with permission. Any reproduction in full or in part This workshop, and the preparation and publication of the report, was made possible through the generous of this publication must credit WWF and ACE support of Peregrine Adventures, which operates expedition-style voyages to the Antarctic Peninsula, the CRC as the copyright owners. Falkland Islands and South Georgia. Peregrine supports WWF-Australia with funds raised through passenger and staff donations on board their ships, the Peregrine Mariner and Peregrine Voyager. The views expressed in this publication are those of the authors and do not necessarily represent Responsible Tourism is a fundamental platform of Peregrine’s operations. In line with this principle of the views of WWF or ACE CRC. sustainability, Peregrine has funded this report in the belief that it provides a foundation for the understanding and long-term conservation of Antarctica’s important habitats and wildlife. Report prepared by: Susie Grant, Andrew Constable, Ben Raymond and Susan Doust. Additional support was provided throughout the project by WWF-Australia and the Antarctic Climate and Additional contributing authors: Roger Hewitt, Harry Ecosystems Cooperative Research Centre. Keys, John Leathwick, Vincent Lyne, Rob Massom, The authors wish to thank all of the participants in the 2006 Experts Workshop, whose expertise and Matt Pinkerton, Ben Sharp and Phil Trathan. thoughtful input provided the material and outcomes for this report, as well as reviews of the manuscript. This report should be cited as: A full list of participants can be found at the end of the report. Grant, S., Constable, A., Raymond, B. and Doust, Cover photo: © Paul Dudley courtesy Australian Government Antarctic Division S. (2006) Bioregionalisation of the Southern Ocean: Report of Experts Workshop, Hobart, September For copies of this report please contact: 2006. WWF-Australia and ACE CRC. WWF-Australia Head Offi ce The CD attached to this report includes appendices, GPO Box 528 analysis tools, data and maps (including a GIS Sydney, NSW, Australia 2001 database) referred to in the report. Tel: +612 9281 5515 wwf.org.au ISBN: 1 921031 16 6 Table of Contents Acknowledgements

Acronyms and abbreviations

Table of Contents

Executive Summary

1. Introduction

1.1 What is bioregionalisation?

Defi ning regions

1.2 Bioregionalisation in the Antarctic context

CCAMLR

Committee for Environmental Protection

1.3 Antarctica and the Southern Ocean

Southern Ocean characteristics

Existing regionalisations for the Southern Ocean

1.4 Experts Workshop

2. Approach to bioregionalisation

2.1 Identifying properties to be captured

2.2 Classifi cation method

Choosing clustering algorithms

2.3 Variables that capture properties

2.4 Uncertainty

3. Physical regionalisation

3.1 Summary of adopted method

Primary regionalisation

Secondary regionalisation

3.2 Results of Southern Ocean bioregionalisation

Primary regionalisation

Uncertainty

Acronyms and abbreviations Secondary regionalisation

ACC Antarctic Circumpolar Current 3.3 Expert review of bioregionalisation results ATCM Antarctic Treaty Consultative Meeting CCAMLR Commission for the Conservation of South Atlantic (Area 48) Antarctic Marine Living Resources Indian Ocean (Area 58) CEP Committee for Environmental Protection CPR Continuous Plankton Recorder Pacifi c Ocean (Area 88) LME Large Marine Ecosystem 4. Future work MPA Marine Protected Area PAR Photosynthetically active radiation 5. Conclusion PF Polar Front List of Appendices (provided on CD) SACCF Southern Antarctic Circumpolar Current Front SAF Subantarctic Front List of workshop participants SC-CAMLR Scientifi c Committee for the Conservation of Glossary of terms Antarctic Marine Living Resources SSH Sea surface height References SST Sea surface temperature STF Subtropical Front

1 Photograph by Wayne Papps, Australian GovernmentDivision © Commonwealth of Australia Antarctic Photograph by Wayne

2 Executive Summary

In September 2006, twenty-three scientists predictable ecosystem properties. The from six countries attended an Experts properties of a given bioregion should differ Workshop on Bioregionalisation of from those of other regions in terms of the Southern Ocean held in Hobart, species composition as well as the attributes Australia. The workshop was hosted by of its physical and ecological habitats. the Antarctic Climate and Ecosystems Classifi cation of regions based only on Cooperative Research Centre, and WWF- biological data is often impractical at larger Australia, and sponsored by Peregrine. scales because of insuffi cient geographic The workshop was designed to assist with coverage, even though there may be the development of methods that might suffi cient data to subdivide smaller-scale be used to partition the Southern Ocean portions of those regions. Physical and for the purposes of large-scale ecological satellite-observed data generally have better modelling, ecosystem-based management, spatial and temporal coverage and greater and consideration of marine protected availability than biological data. These can areas. In 2005, the Commission for the be used to help characterise regions on the Conservation of Antarctic Marine Living basis of environmental properties, physical Resources (CCAMLR) and its Scientifi c processes, primary production, and Committee (SC-CAMLR) considered that habitat type. a bioregionalisation of the Southern Ocean was needed to underpin the development Initial discussions during the workshop of a system of marine protected areas in focused on defi ning the major physical the Convention area. processes in the Southern Ocean, and their relationships with ecological processes. The aim of the workshop was to bring A key aspect of undertaking an together scientifi c experts in their ecologically meaningful regionalisation is independent capacity to develop a ‘proof of to understand how important ecological concept’ for a broad-scale bioregionalisation processes correspond to the physical of the Southern Ocean, using physical and satellite-observed parameters, and environmental data and satellite-measured whether these parameters are appropriate chlorophyll concentration as the primary for use as proxies or surrogates. This may inputs. Work included presentation of depend in part on the end-use application background information, computer-based of the analysis, and the scale at which the analysis undertaken in small groups, and analysis is being undertaken. plenary discussion on the methods, data and results. Workshop participants are Environmental data used as the primary listed at the end of this report. input for analysis during this workshop were chosen based on their spatial At the conclusion of the workshop, a coverage across the Southern Ocean. method had been agreed upon that could The datasets considered included be used to take the bioregionalisation work bathymetry, sea ice concentration and forward. Consensus was achieved on a extent, sea surface temperature, sea surface draft physical regionalisation, and progress height, chlorophyll a concentration, nutrient was made in determining how to include data (silicate, nitrate and phosphate), additional (e.g. biological) data for a more and insolation (photosynthetically complete bioregionalisation. This report active radiation - PAR). outlines the key results of the workshop, and highlights some of the issues discussed. A series of presentations on approaches to bioregionalisation that have been An understanding of the spatial undertaken elsewhere (terrestrial characteristics of large ecosystems such Antarctica, Australia, New Zealand) as the Southern Ocean is important for allowed detailed consideration of the the achievement of a range of scientifi c, relative benefi ts of different methods. conservation and management objectives. The analytical methods used by Lyne and Bioregionalisation is a process that aims Hayes (2005), Leathwick et al. (2006a) to partition a broad spatial area into and Raymond & Constable (2006) were distinct spatial regions, using a range of used as starting points for the analysis environmental and biological information. during the workshop. These methods The process results in a set of bioregions, were refi ned into a single methodology, each with relatively homogeneous and

3 IInn SSeptembereptember 22006,006, ttwenty-threewenty-three scientistsscientists ffromrom sixsix countriescountries attendedattended aann EExpertsxperts WorkshopWorkshop onon BBioregionalisationioregionalisation ooff tthehe SSouthernouthern OceanOcean heldheld inin HHobart,obart, Australia.Australia. TheThe

wworkshoporkshop wwasas hhostedosted bbyy Papps, Australian GovernmentDivision, © Commonwealth of Australia Antarctic Photographer: Wayne tthehe AAntarcticntarctic ClimateClimate aandnd EEcosystemscosystems CCooperativeooperative RResearchesearch Centre,Centre, aandnd WWWF-Australia,WF-Australia, aandnd ssponsoredponsored bbyy PPeregrine.eregrine. TThehe wworkshoporkshop waswas ddesignedesigned ttoo aassistssist wwithith tthehe ddevelopmentevelopment ooff mmethodsethods tthathat mmightight bbee uusedsed toto ppartitionartition tthehe SouthernSouthern OOceancean fforor thethe purposespurposes ooff llarge-scalearge-scale eecologicalcological mmodelling,odelling, ecosystem-ecosystem- bbasedased management,management, andand cconsiderationonsideration ofof marinemarine pprotectedrotected areas.areas. following workshop discussions and including Lyne and Hayes (2005) and practical explorations of the methods. Leathwick et al. (2006a).

Issues examined included the choice of The workshop established a proof of data and extraction of relevant parameters concept for bioregionalisation of the to best capture ecological properties, the Southern Ocean, demonstrating that this analysis can delineate bioregions that agree use of data appropriate for end-user with expert opinion at the broad scale. applications, and the relative utility of Consensus was reached on which of the taking a hierarchical, non-hierarchical, trial bioregionalisations were the most or mixed approach to regionalisation. ecologically and statistically meaningful The fi nal method involved the use of a according to expert opinion. clustering procedure to classify individual The workshop concluded that a statistical, sites into groups that are similar to one hierarchical approach was the most another within a group, and reasonably useful in displaying the different levels dissimilar from one group to the next, of similarity and providing choices on according to a selected set of parameters the degree to which the region might be (e.g. depth, ice coverage, temperature). subdivided on the basis of the chosen This approach shared strong similarities to datasets. The datasets were divided into several previous regionalisation methods, primary and secondary datasets, refl ecting 4 the primary properties of the region and the The secondary datasets used in the analysis Finally, workshop participants discussed secondary environmental properties that were ice concentration and mean chlorophyll priorities for future work, including the might provide smaller-scale subdivisions a values. The addition of these datasets development of further methods to deal to refl ect the spatial heterogeneity of the suggested smaller-scale spatial heterogeneity with uncertainty, understanding of inter- Southern Ocean ecosystem. within the regions particularly in the and intra-annual variation, validation of continental shelf and slope areas, and the results, the incorporation of additional data The primary datasets used in this analysis seasonal ice zone. These results highlighted (particularly biological datasets) and fi ner- were depth, sea surface temperature, the need for further analysis at the scale analysis of particular areas of interest. silicate and nitrate. These highlighted the secondary level. different environmental characteristics of This workshop established a ‘proof of large regions including the continental shelf The fi nal stages of the analysis included concept’ for bioregionalisation of the and slope, frontal features (Subantarctic discussion on how well the defi ned regions Southern Ocean. Continuation of this Front, Polar Front, Southern Antarctic corresponded to our present knowledge work will be an important contribution to Circumpolar Current Front), the deep of the Southern Ocean. Experts provided the achievement of a range of scientifi c, ocean, banks and basins, island groups information on the patterns and features management and conservation objectives, and gyre systems. Other primary datasets that they would expect to see, according to including large-scale ecological current observations and understanding, and that could be usefully considered in future modelling, ecosystem-based management, these largely concurred with the outcomes of analyses were identifi ed by the workshop, and the development of an ecologically the analysis. and included sea surface height representative system of marine and insolation. protected areas. 5 FIGURE 1: Map of Antarctica and the Southern Ocean. (Data from Australian Antarctic Data Centre)

1. Introduction The Southern Ocean covers around 10% and distribution of species and their habitats, Committee for Environmental Protection of the world’s ocean surface, and includes and is an important foundation for efforts (CEP), which reports to the Antarctic some of the most productive marine regions to further understand, conserve and manage Treaty Consultative Meeting (ATCM). on Earth. Although they are among the activities in the marine environment. In 2006, WWF initiated a project to undertake least-studied, the seas around Antarctica are Attempts to classify large ocean areas some of the initial work towards a broad- a critical component of the global climate into meaningful management units have scale bioregionalisation, in partnership system and marine ecosystem. been carried out for coastal and shelf areas with the Antarctic Climate and Ecosystems An understanding of the spatial worldwide, for example in the defi nition Cooperative Research Centre (ACE CRC, characteristics of large ecosystems such of Large Marine Ecosystems (LMEs) Australia) and sponsored by Antarctic as the Southern Ocean is important for (Sherman and Alexander, 1986; Sherman expedition cruise operator, Peregrine. The aim the achievement of a range of scientifi c, and Duda, 1999), Marine Ecoregions of this work has been to develop a ‘proof of management and conservation objectives (Spalding et al., 2006), and the use of LMEs concept’ for a broad-scale bioregionalisation including ecological modelling, together with biogeochemical provinces of the Southern Ocean based on synoptic ecosystem-based management of living (Platt and Sathyendranath, 1988; 1993; environmental data in the fi rst instance. resources, and the establishment of an Longhurst, 1998) to defi ne global regions As part of this work program, an Experts ecologically representative system of for ecosystem-based fi sheries management Workshop was held in Hobart, 4-8 September marine protected areas. (Pauly et al., 2000). 2006, to review and expand upon the Bioregionalisation is a process that aims In 2005, the Commission for the Conservation initial developmental work provided to partition a broad spatial area into of Antarctic Marine Living Resources by Raymond and Constable (2006). distinct spatial regions, using a range of (CCAMLR) and its Scientifi c Committee This document provides a report of environmental and biological information. (SC-CAMLR) identifi ed a series of key tasks that workshop, detailing: The process results in a set of bioregions, each to assist in developing a comprehensive and • background to the workshop and with relatively homogeneous and predictable ecologically representative system of marine bioregionalisation in the Southern Ocean; ecosystem properties. The properties of a protected areas (MPAs) (SC-CAMLR-XXIV, • an agreed approach to bioregionalisation; given bioregion should differ from those of 2005). A broad-scale bioregionalisation • an example of a regionalisation for other regions in terms of species composition of the Southern Ocean was identifi ed as the Southern Ocean based on synoptic as well as the attributes of its physical and an important fi rst step in this process. environmental data; and ecological habitats. Bioregionalisation can CCAMLR agreed that this process will need • future work towards a bioregionalisation assist in providing information on the location to be undertaken in cooperation with the of the Southern Ocean. 6 An understanding of the spatial characteristics of large ecosystems such as the Southern Ocean is important for the achievement of a range of scientifi c, conservation and management objectives. Bioregionalisation is a process that aims to partition a broad spatial area into distinct spatial regions, using a range of environmental and biological information. The process results in a set of bioregions, each with relatively homogeneous and predictable ecosystem properties. The properties of a given bioregion should differ from those of other regions in terms of species composition as well as the attributes of its physical and ecological habitats.

1.1 What is bioregionalisation?

Large ecosystems can be partitioned at a certain species to a particular set of physical range of spatial scales, according to their conditions. However, boundaries may also physical, environmental and biological be gradual, such as in the margins of a characteristics. Variation in climate, desert, where habitats and species from both topography and other physical factors forms the desert and the neighbouring grassland different habitat types, which in turn support gradually blend across a wide transitional different species and communities. Biological area. Transitional areas between adjacent diversity varies throughout this geographic ecosystems, regions or habitats are known space, and may be further infl uenced by as ecotones, and species may be found in factors such as the availability of nutrients decreasing numbers as they reach the edge and food, as well as human activities. of their range. Bioregionalisation provides a simplifi ed interpretation of these physical For example, forests, deserts and grasslands and ecological boundaries. It endeavours to have different physical and environmental separate, say, desert, grassland and forest by attributes, and contain different habitat types drawing boundaries between them such that and communities of species. These different the attributes within each of the bounded regions may occur adjacent to one another; areas are primarily desert, grassland and however, each differs from the others in terms forest respectively. of physical and ecological characteristics. Some species may range across more than one This terrestrial analogy provides a simplifi ed region, whereas others will be more restricted description of the bioregionalisation concept, in their range, according to their ability to and its utility in providing pragmatic solutions live in particular habitat types or ecological to complex ecological problems. Apart from conditions. For example, cacti are uniquely the edges of rocky reefs, regional boundaries adapted to live only in desert conditions, in the oceans are likely to be less sharp (or while certain ubiquitous grasses are found more ‘fuzzy’), and they may be more mobile in parts of the forest and the grassland, as or variable because of the fl uid nature of well as the desert. Migrating birds may travel the marine environment. Regionalisation of across all three regions, while deer inhabit the marine ecosystems is also more complex forest and the grassland but not the desert, and because of their three-dimensional nature. tree-dwelling mammals remain exclusively in However, marine ecosystems can nevertheless the forest. be partitioned using the principles described above to provide a simplifi ed interpretation Boundaries between regions may be sharp, spatial differences in their environmental for example at the interface between a forest characteristics, habitat types and and adjacent alpine areas. Features such as ecological boundaries. the tree-line refl ect the limit of tolerance by

7 Defi ning regions Regions are generally defi ned using a physical and ecological habitats) should which infl uence the characteristics and combination of qualitative (expert opinion, differ from those of adjacent regions. structure of habitats and their associated descriptive data) and quantitative statistical species and communities. An understanding Regions can be defi ned according to the analyses. A range of data on physical, of the spatial extent of different range of species or communities that environmental and biological properties environmental conditions and physical inhabit them. Indicator species may also be can be incorporated into a regionalisation habitats can provide further information used, where individual species are known analysis, according to data availability and on the ecological properties likely to be to exclusively inhabit a certain type of coverage, and specifi c end-use applications. found in each area, and thus the types region. For example, certain species of Statistical procedures for undertaking a of communities or species which might desert snake, grassland lizard and forest regionalisation attempt to partition a broad occur there. As a simplifi ed example, the frog might be used as indicators to defi ne spatial area into discrete regions, each with distribution of freshwater habitats may give these regions. relatively homogeneous and predictable some indication of where frogs are likely to ecosystem properties, but sometimes Alternatively, physical and environmental be found. This is particularly useful where occurring in more than one geographic information can be used to defi ne regions biological information is unavailable. location (Leathwick et al. 2003). The using qualitative methods (e.g. Bailey, Information on the distribution of frogs properties of a given region (both species 1996). Topography, altitude, substratum over a large area may be impractical to composition as well as attributes of the and temperature are among the variables obtain, however freshwater ponds could 8 Photograph by Wayne Papps, Australian GovernmentDivision © Commonwealth of Australia Antarctic Photograph by Wayne

be more easily identifi ed using aerial grassland and desert may be encompassed Clearly, the fi nal regionalisation will be photography. within a much larger unit; for example all dictated by the spatial detail required of these regions would be found in southern Approaches to defi ning regions may also and the specifi c attributes needing to be Africa. Within a particular region, there vary according to the particular application captured in the subdivision. Nevertheless, may also be fi ner-scale ecosystem divisions. of a bioregionalisation analysis. For example, a regionalisation needs to show generally For example, within a forest region, a a manager interested in the conservation how those attributes are nested within the mountain will support different vegetation of reptiles may choose to defi ne regions larger scale heterogeneity of the system. with increasing altitude. Different forest according specifi cally to the distribution of communities may be found higher up the This helps to appreciate whether areas with snakes and lizards, whereas an agricultural mountain, refl ecting changes in topography similar properties but separated in space scientist might be more interested in the and climatic conditions. At an even fi ner may be infl uenced by different external division of regions according to substratum scale, features such as mountain streams, environmental and ecological drivers at type and topography. valleys and rocky outcrops may result in their boundaries. Bioregions may also be defi ned at different different forest communities occurring at spatial scales, according to the biological, the same altitude. Smaller scale ecosystems Approaches to bioregionalisation in the physical or environmental characteristics of or regions can be seen as nested within marine environment have included the interest, and the scale of the data being used ecosystems of a higher order, thus occurring use of physical oceanographic parameters in the analysis (e.g. Bailey, 1996). The forest, within a hierarchical system. (e.g. ocean water masses, fronts, gyres 9 and wave energy), geomorphology (e.g. coverage and greater availability than depth, substratum, sediment characteristics biological data. These can be used to and disturbance regimes), biological help characterise regions on the basis oceanography (e.g. primary and secondary of environmental properties, physical production), fi sh stock distribution and processes, primary production, and abundance (e.g. areas of aggregation and habitat type. fi shing patterns), benthic communities (e.g. An important aspect of undertaking an distribution and community structure) and ecologically meaningful regionalisation marine mammals and birds (e.g. primary is therefore to understand how important feeding and breeding locations). ecological processes correspond to physical Classifi cation of regions based only on parameters, and whether those parameters biological data is often impractical at larger are appropriate for use as proxies or scales because of insuffi cient geographic surrogates. This may not require much coverage, even though there may be ecological detail in the fi rst instance, suffi cient data to subdivide smaller-scale since physical and environmental data can portions of those regions (Belbin, 1993). provide an understanding of environmental Physical and satellite-observed data heterogeneity which will inevitably affect generally have better spatial and temporal the ecology of a region. Photograph © Roger Kirkwood, courtesy Australian GovernmentDivision Antarctic

1.2 Bioregionalisation in the Antarctic context

Bioregionalisation of the Southern Ocean has respectively, and each is further divided • determination of statistical analyses relevance in a variety of applications within into statistical subareas for catch reporting required to facilitate a bioregionalisation, different scientifi c fi elds and for conservation and management purposes (see Figure 1). including use of empirical, model and and management across the Antarctic Statistical subareas were defi ned on the expert data; Treaty System. An understanding of spatial basis of ocean characteristics, fi sh stock • development of a broad-scale ecosystem characteristics is necessary to distributions and the location of fi shing bioregionalisation of the Southern achieve a range of objectives in the Antarctic activities (Everson, 1977; Kock, 2000), Ocean, based on existing datasets; and context, including: thus providing one example of an existing • delineation of fi ne-scale provinces bioregionalisation of the Southern Ocean. within regions, where possible. • ecosystem modelling; Subareas are used in catch reporting, and • ecosystem-based management of marine As part of this ongoing work, CCAMLR enable the implementation of conservation living resources; will hold a workshop in 2007 with the aim and management measures regionally or • effective and systematic planning and of providing advice on a bioregionalisation for individual stocks. management of other human activities; of the Southern Ocean, including, • identifi cation of biodiversity units and The primary objective of CCAMLR is the where possible, advice on smaller-scale areas of high conservation value; conservation of Antarctic marine living delineation of provinces and potential areas • establishment of a comprehensive and resources, where conservation includes for protection to further the conservation ecologically representative system of rational use. CCAMLR has pioneered objective of CCAMLR. This workshop MPAs; and a precautionary, ecosystem approach to will involve members of both the • directing further research. marine living resource management, and CCAMLR Scientifi c Committee and defi nes the Antarctic marine ecosystem as the CEP, as well as external experts Recent discussions within the CCAMLR “the complex of relationships of Antarctic (SC-CAMLR-XXIV, 2005). Scientifi c Committee (SC-CAMLR- marine living resources with each other XXIV) and the Committee for Committee for and with their physical environment”.1 Environmental Protection (CEP IX) have Environmental Protection agreed the importance of undertaking a In 2005, the CCAMLR Workshop on The Antarctic Treaty and its Protocol bioregionalisation of the Southern Ocean, MPAs considered the scientifi c work on Environmental Protection apply to and highlighted the need to work together required for development of a system the area south of 60°S, thus covering in achieving this common objective. of protected areas to assist CCAMLR a smaller marine area than CCAMLR. in achieving its conservation objectives. The Environmental Protocol deals CCAMLR A broad-scale bioregionalisation of the with environmental impact assessment, The Convention on the Conservation Southern Ocean was identifi ed as an conservation of Antarctic fl ora and fauna, of Antarctic Marine Living Resources important fi rst step in this process waste disposal and management, prevention (CCAMLR) applies to all marine living (SC-CAMLR-XXIV, 2005). of marine pollution, and area protection resources within the area south of a line and management. The Committee for approximating to the Polar Front. The CCAMLR has identifi ed a series of key tasks Environmental Protection (CEP) provides Convention Area is divided into three sectors to be undertaken towards bioregionalisation: advice and recommendations to the ATCM corresponding to the adjacent Atlantic, • collation of existing data, including in connection with the implementation of Indian and Pacifi c oceans. These sectors are benthic and pelagic features and the Environmental Protocol. referred to as Statistical Areas 48, 58 and 88 processes; 10 1.3 Antarctica and the Southern Ocean Annex V of the Environmental Protocol states that Parties shall seek to identify a Southern Ocean characteristics series of Antarctic Specially Protected Areas The Southern Ocean extends across a total 2 (ASPAs) (including marine and terrestrial area of almost 35 million km , and consists areas) within a ‘systematic environmental- of distinct provinces that differ physically geographic framework’2. This term has and chemically (e.g. temperature, sea been defi ned as: “a method of classifying ice, nutrients and currents), as well as or organising subsets of environmental ecologically. It is characterised by deep and geographic characteristics such as basins, separated by large, mid-oceanic different types of ecosystems, habitat, ridges and containing prominent plateaus geographic area, terrain, topography, and island groups. climate, individual features and human Two major currents dominate the Southern presence into geographic regions. Each Ocean system. The Antarctic Circumpolar region would be distinctive or in some way Current (ACC) (or “West Wind Drift”) fl ows different from other regions but some might eastwards around the continent, driven by the have characteristics in common.” (ATCM prevailing westerly winds. The ACC forms XXIV/WP012, 2001). a unique link connecting all of the world’s The bioregionalisation work proposed by major oceans through an unbroken water CCAMLR corresponds closely to current mass surrounding the Antarctic continent, efforts by the CEP to elaborate a systematic (Orsi et al., 1995). However, its path is environmental-geographic framework, in infl uenced by topographic features such as the particular through the terrestrial Antarctic Kerguelen Plateau and the Scotia Arc, which Environmental Domains Analysis being defl ect fronts and generate eddies. Closer to undertaken by New Zealand for the Antarctic the continent, easterly winds form a series of continent (Morgan et al., 2005). clockwise gyres (the largest of these being in the Ross Sea and Weddell Sea) that combine In outlining the work programme for to form the westward fl owing Antarctic bioregionalisation, CCAMLR recognised Coastal Current, also known as the “East the relative expertise of the CEP, and Wind Drift”. suggested that the CEP should be invited to undertake the initial work necessary to The Subtropical Front (STF) marks the develop a bioregionalisation of the coastal northernmost extent of the ACC, separating provinces, as an extension of its terrestrial warmer, more saline subtropical waters bioregionalisation work. At its meeting in from fresher, cooler subantarctic surface 2006, the CEP undertook to engage fully waters (Orsi et al., 1995). Further south, with CCAMLR on this work, and agreed the majority of ACC water is transported in on the importance of such an analysis the Subantarctic Front (SAF), and also in 1 CCAMLR, Article 1 the Polar Front, which marks the transition in contributing to its conservation and 2 Protocol on Environmental management objectives (CEP IX, 2006). Protection, Annex V, Article 3(2) to very cold and relatively fresh Antarctic 11 Surface Water, and separates Southern Ocean waters from the Atlantic, Pacifi c and Indian oceans to the north. The Polar Front also marks the northerly limit of many non-migrating Antarctic species (Knox, 1994), including Antarctic krill (Euphausia superba), the staple food of many of the Southern Ocean’s seabirds, marine mammals and fi sh. Closer to the Antarctic continent, upwelling of very dense, cold abyssal waters occurs at the southern boundary of the ACC.

The Southern Ocean plays an important role in the global ocean circulation system. Figure 2 shows the relationships between the frontal systems and the greater patterns of ocean circulation. Of note are the extremely cold winds blowing off the Antarctic Ice Sheet which cool the coastal waters. In certain recurrent locations (coastal polynyas), these create high rates of sea ice formation. This in turn leads to the formation of cold, dense saline water that sinks to form Antarctic Bottom Water. FIGURE 2: Three-dimensional structure of water masses, showing relationship between This complex system also sees a continual the ACC and deep water (Figure reprinted with permission from: Rintoul, 2000). surface expression in the Southern Ocean of the global ocean’s normally deep nutrient polynyas (large areas of open water) occur have experienced a statistically signifi cant layer, which is the primary reason for the around the continent (Arrigo and van Dijken, decreasing trend in sea ice areal extent since sustained high productivity in the region. 2003), and there are also two deep water 1978 (see inset in Figure 3, from Kwok and In the tropics, this nutrient layer only polynyas in the Weddell and Cosmonaut seas Comiso, 2002). Recent results imply that this reaches the surface through upwelling. (Morales Maqueda et al., 2004). Polynyas change may result from changes in dynamic The continental shelf surrounding Antarctica constitute major regional sea ice “factories”, (i.e., wind-driven) forcing (Massom et al., is unusually deep compared to elsewhere sites of major water-mass modifi cation and, 2006). These factors, combined with the in the world, as a result of scouring by ice in places, enhanced biological activity. profound impact of the Antarctic Peninsula shelves and crustal depression caused by The seasonal cycle of sea ice advance as a meridional blocking feature that extends the weight of continental ice (Clarke, 1996). and retreat is one of the major drivers of to low latitudes and oceanic characteristics, The continental shelf is generally narrow, physical and ecological processes in the suggest that the WAP region should be treated except in large embayments such as the Southern Ocean. On the hemispheric scale, as a separate regime. Ross Sea and Weddell Sea. the sea ice cover in winter interacts with Forming an important habitat for a wide The Southern Ocean is covered by a band key oceanic and biological boundaries such range of organisms specifi cally adapted to its of largely seasonal sea ice that extends as the continental shelf break, the southern presence, sea ice plays a dominant defi ning from a maximum southerly extent of ~75oS boundary of the Antarctic Circumpolar role in structuring high-latitude marine Current (Tynan, 1998) and the Antarctic northwards as far as ~55oS at its maximum ecosystems (Ackley and Sullivan, 1994; Divergence, the latter being an important extent. The width of this band is highly Brierley and Thomas, 2002; Eicken, 1992; zone of upwelling. The areal extent of variable, ranging from a few hundred Lizotte and Arrigo, 1998; Nicol and Allison, Antarctic sea ice varies annually by a factor kilometres in the Indian Ocean sector to 1997), and on a variety of scales. The most of ~5, from a maximum of 18-20 x 106 km2 ~1600 km in the Weddell Sea. Given the productive areas of the Southern Ocean lie in in September-October to 3-4 x 106 km2 each relatively narrow width of the continental the Seasonal Ice Zone, between the maximum February. As such, it is predominantly a shelf surrounding the Antarctic continent, northern extents of sea ice in winter and seasonal sea ice zone, although large regions a large proportion of the Antarctic ice cover summer. Here in particular, Antarctic krill of perennial ice persist in the western occurs over deep ocean, where it is exposed and other planktonic organisms support an Weddell Sea, Amundsen Sea and Ross Sea to a zone of strong cyclone activity and ocean abundance of fi sh, birds, seals and whales. In and southwest Pacifi c Ocean though summer waves and swell. The latter create a well- addition, the ice edge is typically a region of (Gloersen et al., 1992). developed circumpolar marginal ice zone, enhanced biological activity during the melt which effectively protects the inner pack The major features driving the dynamics season in particular (Nicol and Allison, 1997; from incoming ocean wave energy. of sea ice are shown in Figure 3. Smith and Nelson, 1986; Smith et al., 1988; Sullivan et al., 1993). The coastal zone is complex, with sea ice The Antarctic Peninsula region is the only distribution and characteristics (of both Antarctic sector to have experienced a rapid Although in the past characterised as simple, pack and fast ice) being affected by coastal warming trend over the past 50 years, of the Antarctic food web involves complex confi guration and the presence of grounded ~0.5oC per decade (Vaughan et al., 2001). relationships between primary producers and icebergs, which are in turn closely linked to Moreover, the West Antarctic Peninsula higher predators, as well as abiotic factors. bathymetry (Massom et al., 2001). Coastal (WAP) region is the only Antarctic sector to The Antarctic ecosystem is characterised by 12 FIGURE 3: Map of climatological (mean) satellite-derived sea ice motion for 1997 (courtesy US National Snow and Ice Data Center; Fowler, 2003), with broad-scale sea ice sectors superimposed. Explanations are provided in the fi gure. The motion vectors are projected to a 25 x 25 km resolution grid. Dominant features in the climatological ice drift pattern are 3 major ocean gyre systems, the westward-drifting Antarctic Coastal Current and eastward fl owing Antarctic Circumpolar Current, with regions of retrofl ection associated with the ocean bathymetry i.e., ocean bathymetric “steering”. Figure by R. Massom. strong seasonal cycles and major food-web differences between the benthic faunas of East differences that are intimately related to the and West Antarctica (Clarke and Johnston, annual sea ice growth-decay cycle and sea 2003). A recent study on the biodiversity ice conditions, as well as associated ocean and biogeography of subantarctic mollusca dynamics (mixing), water density and nutrient (Linse et al., 2006), using species from the availability (Garrison and Mathot, 1996; continental shelf areas (0-1000 m), identifi ed Legendre et al., 1992; Lizotte, 2001). the following distinct sub-regions in the Existing regionalisations for the Southern Ocean: Antarctic Peninsula, Weddell Southern Ocean Sea, Dronning Maud Land, Enderby Land, The Southern Ocean has been divided into Wilkes Land, Ross Sea, and the independent large-scale regions before, primarily based Scotia arc and subAntarctic islands (Figure 4). on physical characteristics such as frontal These divisions have also been used by WWF features (Orsi et al., 1995; Longhurst, 1998) and The Nature Conservancy (TNC) in a and ice dynamics (Tréguer and Jacques, study to synthesise existing classifi cations into 1992). Information on the distribution of a system of Marine Ecoregions of the World species has been used in biogeographic (Spalding et al., 2006). classifi cations of benthic fauna (Ekman, Tréguer and Jacques (1992) defi ned fi ve 1953; Hedgpeth, 1970; Dell, 1972), and also functional units south of the Polar Front on by CCAMLR in the defi nition of statistical the basis of ice and nutrient dynamics. This subareas on the basis of fi sh stock distribution work demonstrated the role of ice dynamics (Everson, 1977) (see Figure 1). in controlling phytoplankton initiation In the southern Indian Ocean, some smaller and growth, and the nutrient regimes that scale regionalisations have been attempted discriminate each of these units. Defi ned in the development of a bioregionalisation in units include the Polar Front Zone, located Australian waters to assist in regional marine between approximately 60°S and 55°S, planning (Lyne et al. 2005), the designation of and the Permanently Open Ocean Zone marine reserves (Meyer et al. 2000), which lies between the Polar Front and the and benthic habitat mapping (Beaman maximum northern extent of winter sea ice. and Harris, 2005). The Seasonal Ice Zone is located between the northern limits of the pack-ice in winter Early biogeographic classifi cations for and in summer, while the Coastal and the Southern Ocean delineated large-scale Continental Shelf Zone is adjacent to the provinces according to the distribution of Antarctic continent. The Permanent Ice Zone benthic fauna (Hedgpeth, 1970; Dell, 1972). incorporates ocean areas under ice shelves. More recent studies have largely confi rmed these broad-scale patterns regions, although Orsi et al. (1995) described large-scale frontal there are now thought to be signifi cant features of the ACC, based on historical 13 hydrographic data. Gradients in ocean surface properties were used to defi ne three major fronts within the ACC which separate water masses and fl ow characteristics. These are shown in Figure 5.

Longhurst (1998) proposed a global system of ocean classifi cation based on a simple set of environmental variables (sea surface temperature, mixed layer depth, nutrient dynamics and circulation) together with planktonic algal ecology. In this classifi cation scheme (Figure 6) the Southern Ocean includes two provinces in the Westerly Winds Biome between approximately 40oS and 50oS (South Subtropical Convergence Province and SubAntarctic Water Ring Province) and two in the Antarctic Polar Biome between 50oS and the continental coast (Antarctic Province and Austral Polar Province).

The LME classifi cation system defi nes the Southern Ocean as a single unit (Sherman and Duda, 1999), while several other classifi cations defi ne only a small number of concentric rings around the continent. However, the Southern Ocean has a variety of distinct provinces within these larger regions which differ in their chemical, physical and ecological characteristics, and which show considerable longitudinal, as well as latitudinal variation. Improved data coverage and availability through satellite imaging, and improved understanding of ocean characteristics through ecosystem modelling makes it now possible to elaborate on these FIGURE 4: Biogeographic areas of the Southern Ocean defi ned by Linse et al. (2006), using distribution records previous regionalisations using a wider range for shelf (0-1000 m) species of shelled gastropods and bivalves and broader coverage of data. (Figure reprinted with permission from: Linse et al, 2006)

30∞W 0∞ 30∞E

STF

SAF

PF sACC dary

C boun sAC

40∞S 40∞S

90∞W 90∞E

40∞S 40∞S

0500 1000 2000 Km

Projection: Polar stereographic True scale at 71∞S

150∞W 180∞ 150∞E

FIGURE 5: Fronts of the Southern Ocean, as FIGURE 6: Classifi cation of the Southern Ocean, Longhurst (1998) defi ned by Orsi et al. (1995) (Reprinted from: Ecological Geography of the Sea, A.R. Longhurst. Copyright (1998), with permission from Elsevier) 14 1.4 Experts Workshop

The aim of the Experts Workshop was to At the start of the workshop, background review the methods for identifying major presentations were given on some of the provinces, collate available synoptic datasets, major physical processes in the Southern and to gain input and recommendations from Ocean, and initial discussion focused experts on the process and the results. In on the relationships between physical particular, the workshop aimed to develop and ecological processes. A series of a “proof of concept” for a broad-scale presentations were also given on approaches bioregionalisation of the Southern Ocean, to bioregionalisation that have been using physical and environmental data as undertaken elsewhere, which allowed the primary input. detailed consideration of the application of different methods. A list of the workshop participants is provided at the end of this report. Participants then investigated different aspects of data analysis and refi nement of Specifi c objectives of the workshop methods in small groups, focusing initially were to: on their regions of particular expertise • review and assess the processes (e.g. South Atlantic, East Antarctica, Ross developed to date and the proposed Sea) and later looking at the Southern methods; Ocean as a whole. Selected physical • discuss and make recommendations on datasets were provided for use in the data types to be included in a broad-scale initial analysis, and others were made bioregionalisation; available by participants during the week. • collate appropriate datasets; The analytical methods used by Lyne and • apply the approved method(s) to the Hayes (2005), Leathwick et al. (2006a) and Southern Ocean using available datasets, Raymond and Constable (2006) were used to test and validate the process and as starting points for the analysis during the produce a ‘proof of concept’ including workshop. These methods were refi ned into maps of the defi ned broad-scale a single methodology, following workshop provinces; discussions and practical explorations of the • assess preliminary results and broad-scale methods. Appendix I gives further details provinces, given present knowledge of on the background and technical aspects of the Southern Ocean. each of these methods. • provide recommendations on products to be developed, including the fi nal report, The fi nal stages of the workshop included maps, illustrations, datasets and a GIS discussion on how well the defi ned regions (or other) database; and corresponded to our present knowledge • provide recommendations on datasets of the Southern Ocean. Priorities were and/or method(s) that might be used identifi ed for further work on issues to develop further fi ne-scale including uncertainty, understanding of bioregionalisations. inter- and intra-annual variation, validation of results, the use of additional data The workshop was held over fi ve days, and (particularly biological datasets) and included background presentations, plenary fi ner-scale analysis of particular areas discussion, and computer-based analysis in of interest. small groups.

The Southern Ocean covers around 10% of the world’s ocean surface, and includes some of the most productive marine regions on Earth. Although they are among the least- studied, the seas around Antarctica are a critical component of the global climate system and marine ecosystem.

15 2. Approach to bioregionalisation

This section describes the approach to bioregionalisation that was used as a starting point for the workshop discussions and analysis. Descriptions of each step are presented here, together with background information on issues that must be considered. Further technical detail is provided in Appendix II. A summary of the fi nal method adopted is presented in Section 3.

The regionalisation process can be partitioned into the following steps:

1. Identify the ecological patterns and processes that have relevance to the end- use application of the regionalisation

2. Identify the major environmental drivers or properties that control these patterns and processes, and extract relevant parameters describing those properties

3. Pre-process the data (e.g. normalise, transform, smooth)

4. Compile a data matrix of individual sites (rows) by properties (columns)

5. Apply a clustering procedure to group sites with similar properties

6. Post-process the clusters to meet any application-specifi c constraints on the regions (e.g. minimum size)

7. Expert review of the regions to ensure suitability for the application.

This process can be iterative. Ideally, the initial process will establish the mechanisms by which new data and/or knowledge could be incorporated into revisions of the bioregionalisation, although this would be expected to assist more in establishing or revising smaller scale subdivisions rather than altering the higher level bioregionalisation.

Figure 7 is a schematic representation of the bioregionalisation process, illustrating how data selected to refl ect ecological processes can be used to defi ne bioregions.

16 © WWF-Canon, Sylvia RUBLI Relevant parameters extracted

Data transformed and normalised Image: European Space Agency (P. Carril) Space Agency (P. Image: European Relevant ecological processes identifi ed Gridded datasets created Appropriate synoptic environmental data obtained via satellite and large-scale datasets

Sea surface 30∞W 0∞ 30∞E temperature Several datasets overlaid

Bathymetry 40∞S 40∞S

90∞W 90∞E

Nitrate 40∞S 40∞S concentration

0500 1000 2000 Km

Projection: Polar stereographic True scale at 71∞S One pixel (or site) has a Data matrix created 150∞W 180∞ 150∞E particular combination of for all pixels (sites) temperature, nitrate and bathymetry properties PIXEL SST DEPTH NOX 1 1.2 113 2050 2 2.3 203 1280 3 0.4 181 550

Clustering procedure groups pixels with similar properties

Final regionalisation

FIGURE 7: Schematic representation of the bioregionalisation process

17 2.1 Identifying properties to be captured

An important fi rst step in the structure and function, which primarily would bioregionalisation analysis is to identify subdivide areas according to the magnitude distinct ecological processes and their of productivity and its predictability in time. defi ning properties. The identifi cation of Further subdivision would relate to the ecological processes to be captured in the diversity of habitats and the relationships regionalisation is likely to be driven by of species and food-webs to those habitats. the requirements of a particular The process will need to differentiate end-use application. between areas with relatively constant features from those that are highly variable, Ideally, a bioregionalisation would delineate even though they may have similar mean units that, depending on the scale, clearly values. This is because a region with a large separate habitats, communities and amount of disturbance can accommodate ecosystems. In this ideal world, populations different assemblages involving opportunistic would reside wholly within these areas. species as well as those that require long-term In reality, there is considerable complexity stability. Some areas in a bioregionalisation that needs to be addressed because of the may need to represent large areas of habitat different relationships that species have with discontinuity or disturbance, which could the environment and other biota (Andrewartha be important regions in themselves. As a & Birch 1984). A bioregionalisation aims result of these considerations, a goal for to capture the properties of the important a bioregionalisation is to capture not only relationships rather than, necessarily, simply the differences in diversity and the suite of trying to circumscribe the distributions of ecosystem relationships, but also the potential whole populations of species. differences in environmental stability. This concept is illustrated in Figure 8. Some species will be found closely aligned with FIGURE 8(a): distribution of species with environmental gradients. Other species respect to a single environmental gradient, will appear in areas with high levels of say temperature. Each polygon represents perturbation, such that environmental factors a different taxon, colours represent similar are mixed and ever changing in their relative taxa. Width of bars represents relative distributions. Yet others will exploit the abundance for a taxon. diversity of patches in fringing habitats and ecotones. For mobile species, some taxa will be found across most areas but only some areas will be important to them as feeding or ENVIRONMENTAL GRADIENT 1 reproductive areas. An important step in the process is to determine how to accommodate FIGURE 8(b): as for (a) but with environmental gradients and overlaps in distributions relative to two environmental the regionalisation. gradients - distributions on gradient 1 are as in (a). A general discontinuity between The marine environment comprises species distributions in (a) might be roughly three dimensions – geographic space and at the midpoint (green on the environmental gradient). In (b) the separation between depth. Distribution of biota in the pelagic groups of taxa becomes clearer on the environment is mostly determined by the second gradient. potential productivity in the water masses and the movement of those water masses in

space and depth. The benthic environment GRADIENT 2 ENVIRONMENTAL has additional features refl ecting variation ENVIRONMENTAL GRADIENT 1 in the depth, substratum types and roughness of the seafl oor, and the degree to which FIGURE 8(c): The environmental conditions this promotes interaction with the pelagic form a patchwork in geographic space which realm. These features are often considered can result in regions of similar environmental to the primary drivers of environmental conditions being separated geographically. Here, parallelograms represent spatial areas heterogeneity. Secondary drivers are more LATITUDE with colours representing two environmental ephemeral or changing over time. In the parameters (outer colour is from environmental Antarctic, they would also include other gradient 1 and inner colour is from gradient features of the environment such as the 2). Each area is expected to have a species annual cycle of advance and retreat of the composition consistent with that composition LONGITUDE at the intersection of two respective sea ice zone. environmental gradient colours in (b). A bioregionalisation would generally try to represent the heterogeneity in ecosystem FIGURE 8: Conceptual diagram illustrating potential relationships between species along gradients of environmental parameters. 18 Photograph by Brian Ball, Australian GovernmentDivision © Commonwealth of Australia Antarctic

2.2 Classifi cation method

The intent of a regionalisation is to Those groupings (which are calculated in are formed from those clusters. A cluster partition the study area into a set of discrete ‘environmental space’, i.e. based only on is a group of sites that are considered to spatial regions, each with relatively environmental properties, and ignoring have similar environmental properties. homogeneous ecosystem properties. In the spatial information) are then projected back However, because the clustering process is bioregionalisation analysis, regions are into geographic space in order to fi nd the based on environmental similarity (and not selected by grouping sites with particular spatial extents of the resulting regions. Thus, spatial information), a single cluster may characteristics. In some cases there may be the regions are discrete in environmental contain sites that are spatially separated. specifi c, known characteristics that can be space, but may be scattered or fragmented in A region is thus considered to be a group used to delineate region boundaries, such as geographical space (i.e. there may be several of sites that belong to the same cluster, but water temperature changes across oceanic regions with the same properties located in which also form a contiguous spatial area. fronts. Another example is to separate the different geographic areas). A single cluster may produce a number of continental shelf from the continental slope Choosing clustering algorithms regions, each of which have the same general by choosing an appropriate bathymetric There are a large number of clustering properties, but which are spatially distinct. contour, say 1000 m. Generally, however, algorithms that could potentially be used, Clustering algorithms are often based around the expectation is that the regions will refl ect all of which have assumptions or limitations the concept of a dissimilarity metric, which a natural clustering of the environmental or that may preclude their use in particular (in the context of a regionalisation) is used biotic data. circumstances or with particular types of data. to calculate how dissimilar two sites are, Clustering algorithms are well suited to Thus, the outcomes of the bioregionalisation given their ecosystem properties (physical bioregionalisation analysis, as they are could be infl uenced by the choice of the or biological data). The clustering of sites designed to partition a large data set into algorithm. The aim is to develop a clustering into regions is carried out in such a way that a number of subsets, each with relatively process that is consistent with the data and the intra-region dissimilarity of sites is low similar properties that differ from those for which the results are likely not to change (i.e. sites within a region are similar to each much with alternative clustering algorithms. of the other subsets. In the context of a other) relative to inter-region dissimilarities. Consideration will need to be given, inter alia, regionalisation, the clustering process takes Dissimilarity-based clustering methods can to algorithm assumptions, complexity, sites (or cells) from a grid in geographic be broadly divided into hierarchical or non- and accuracy. space. Each site has associated ecological hierarchical schemes. Further information properties (physical and/or biotic data) and It is important to make the distinction on these schemes, and the issues related to this information is used to group together sites between the clusters that are produced by selecting clustering algorithms, is provided with relatively similar ecological properties. a clustering algorithm, and the regions that in Appendix II. 19 a similar scale, say 0 to 1, while preserving the rates of change between different levels of the variable that need to be maintained in the analysis. Alternatively variables might be transformed where biological changes are greater in one part of the gradient, e.g., a log transformation might be used with ocean depth, given that rates of biological turnover are rapid near the ocean surface, but decrease with progression to deeper waters. Care needs to be taken to ensure that the properties of the variables and their relationships to other variables are not altered in the process. Variables that infl uence multiple ecological FIGURE 9: Water mass profi le of nitrate concentration from Antarctica processes will refl ect different aspects of to the tropics, showing surface those processes depending on how they are expression of the nutrient layer in incorporated into the analyses. the Southern Ocean. (Figure from http://woceSOatlas.tamu.edu. Orsi The spatial and temporal scales of the data and Whitworth, 2005) should be appropriate to the desired scale of the areas. Data with fi ne-scale spatial or 2.3 Variables that capture properties temporal structure may need to be smoothed for use in broad-scale regionalisations. For Once the relevant physical properties have algorithms may impose some constraints on pelagic applications, the selection of spatial been identifi ed, appropriate data must be the types of data that can be used. regions is complicated by the depth structure selected to be included in the analysis. The Figure 9 demonstrates how a particular of the water column and temporal variability data that are used in the clustering method variable (nitrate concentration) can capture at seasonal and longer timescales. A should be matched to the ecological patterns environmental properties (surface expression hierarchical approach is often used to assist and processes and spatial and temporal of nutrients) across a broad spatial area. in resolving problems of scale. The levels scales that are important to the end-use of the hierarchy can represent either spatial application. However, there is considerable Once relevant data have been collated, the scales, or different ecological processes. latitude for choice within this broad study area is divided into a grid of sites A process-oriented hierarchy often has an guideline. Importantly, the data used in the (or interpolated from point observations), approximate spatial structure due to the clustering procedure will not necessarily be at a suffi ciently fi ne spatial scale to enable spatial scales of the processes. the raw observations from fi eld sampling. appropriate resolution of the fi nal areas. To illustrate the concept of temporal The data may be transformed or be analysed Descriptive statistics – such as means, variability at seasonal timescales, Figure within a model (processing algorithm) to variances, and other ecologically relevant 10 shows the mean monthly chlorophyll provide the necessary inputs to clustering. information, including rates of change of a concentrations for each month during For example, ice concentration maps can parameters – are computed from the input summer (December to March). The be routinely obtained from satellite passive data at a site level. Site data may be further seasonal variability of this property must be microwave data (since 1978). However, daily processed if necessary. This might include taken into account when using such data to ice concentration or mean concentration spatial or temporal smoothing of the data capture ecological processes. Nevertheless, over time might not alone be indicative of in order to ensure that the data provides the ecological processes of importance. The Figure 10 shows that certain areas maintain information at an appropriate scale for the amount of time an area is free of signifi cant high levels of chlorophyll a concentration regionalisation. The algorithms for selecting concentrations of sea ice over the course of throughout this period, and thus a summer areas often also require data to be normalised a year may be more important in terms of mean value may be appropriate for use in a so that variables with different measurement productivity in an area or the amount of time bioregionalisation analysis. units can be statistically combined. the area might be open to feeding activity Further information on scaling and weighting of birds, seals and whales. Data availability Variables of comparable type but measured of variables is provided in Appendix II. and the choices of subsequent processing on different scales are often normalised to

DECEMBER JANUARY FEBRUARY MARCH

FIGURE 10: Mean monthly chlorophyll a concentrations for each month during summer (Dec-Mar) (Images provided by the SeaWiFS Project, NASA/Goddard Space Flight Center and ORBIMAGE) 20 2.4 Uncertainty

A regionalisation requires an assessment of the uncertainties in the locations of the boundaries between areas. In addition, as assessment should be made of whether the heterogeneity within an area is not suffi ciently great that the area should not be differentiated from one or more of its neighbours. Here, the term “uncertainty” is used to describe the effects of a number of different processes, including imprecision in data (for example measurement error, and bias due to incomplete or unbalanced observations), model uncertainty (uncertainty within models that have been used to derive one variable from others, such as climatologies or primary productivity models), and epistemic uncertainty (lack of knowledge of how to go about the regionalisation process; Raymond and Constable 2006). Each of these can affect the resulting region boundaries. Note that stochastic, seasonal, or other temporal or spatial variability in data represents the temporal or spatial variability of the underlying ecosystem processes, and is not treated as uncertainty. However, if it is not clear how this variability should be incorporated into the regionalisation (e.g. should summer or annual means be used?) then this would, in turn, be a source of uncertainty. A key output of an uncertainty analysis would be an assessment of the uncertainty in region boundary locations. This would indicate to end-users where they might expect the region boundaries to change if the data or analysis methods were to be updated or changed.

21 Photograph by John Kelly, Australian GovernmentDivision © Commonwealth of Australia Antarctic Photograph by John Kelly, 22 Photograph by Grant Dixon, Australian GovernmentDivision © Commonwealth of Australia Antarctic 3. Physical regionalisation 3.1 Summary of by bathymetric data) was included because adopted method of the clear ecological differentiation of the shelf, slope and abyssal regions as well The classifi cation method adopted as its infl uence on upwelling, eddying and during the workshop was a mixed non- as a potential source of iron. Bathymetry hierarchical and hierarchical approach. The (Figure 12) was transformed (log10) to classifi cations were performed on a 1/8th give most weight to the shallower areas less degree grid, covering the marine area from than 2500 m with a greater opportunity to 80°S to 40°S. The full set of 720,835 grid differentiate the shelf break and slope. cells was subjected to a non-hierarchical clustering to produce 40 clusters. The mean Silicate and nitrate concentrations data values for each of the 40 clusters was (Figures 13 and 14) were included calculated and a hierarchical classifi cation to provide information on nutrient was then performed to produce a characteristics. Silicate concentration dendrogram and the fi nal clustering. also provides a measure of actual primary production (particularly in diatom- Sites with missing data were excluded from dominated areas), since silicate is taken the analyses. These were principally sites up during photosynthesis in the production shallower than 200 m depth, for which the of diatom shells. The silicate layer was chosen nutrient data did not apply. These found to be particularly useful for accurately excluded sites are shown in the maps as differentiating water masses refl ecting white. Future work will need to fi ll in these plankton communities in deeper water and missing cells, for example by considering along the various fronts. The nitrate and their other attributes. silicate climatologies at the 200 m depth Primary regionalisation layer were used rather than the surface The primary regionalisation used the layer as this is a better indicator of available following datasets: nutrients, whereas surface nutrients are • bathymetry (log10 transformed) likely to be depleted in areas of nutrient- • sea surface temperature (SST) limited productivity. However, the use of the • nitrate (NOx) concentration 200 m depth layer resulted in missing data in • silicate (Si) concentration the shelf areas of less than 200 m depth.

Descriptions of each of these datasets are Sea surface height (SSH) and insolation provided in Appendix III. (mean summer climatology of photosynthetically-active radiation (PAR) The workshop agreed that the ocean water at the ocean surface) were considered as masses combined with topography of additional primary variables that would the ocean fl oor were likely to defi ne the have utility in defi ning frontal systems and primary features of the Southern Ocean productivity respectively, however they were and coastal Antarctic systems. Sea surface not used at this stage because of insuffi cient temperature was included as a proxy for time, and because the currently available the different water masses of the Southern datasets were incomplete. These datasets Ocean (Figure 11). Topography (captured should be considered in future analyses.

Physical environmental data used as the input for analysis during the workshop were chosen based on their spatial coverage across the Southern Ocean. The datasets considered included bathymetry, sea ice concentration and extent, sea surface temperature, sea surface height, chlorophyll a concentration, nutrient data (silicate, nitrate and phosphate), and insolation (photosynthetically active radiation - PAR).

23 FIGURE 11: Mean annual sea surface temperature (SST). Monthly values from NOAA Pathfi nder satellite annual climatology, averaged over the period 1985-1997 (Casey and Cornillon 1999)

Temperature (°C)

-1.95 - -0.13 -0.12 - 2.01 2.02 - 4.46 4.47 - 6.83 6.84 - 9.05 9.06 - 11.18 11.19 - 13.40 13.41 - 18.22

FIGURE 12: Bathymetry of the Southern Ocean. Depth data from the GEBCO digital atlas (IOC, IHO and BODC, 2003).

150oW Depth (metres)

0 - 500 500 - 1000 1000 - 1500 1500 - 2000 2000 - 2500 2500 - 3000 3000 - 4000 4000 - 5000 > 5000

24 FIGURE 13: Silicate concentration (at 200 m depth). Climatology from the WOCE global hydrographic climatology (Gouretski and Koltermann, 2004)

Si (umol/kg)

2.47 - 4.58 4.59 - 7.96 7.97 - 18.08 18.09 - 38.76 38.77 - 64.93 64.94 - 76.74 76.75 - 86.87 86.88 - 110.08

FIGURE 14: Nitrate concentration (at 200 m depth). Climatology from the WOCE global hydrographic climatology (Gouretski and Koltermann, 2004)

NOx (umol/kg)

5.55 - 12.40 12.41 - 15.58 15.59 - 18.65 18.66 - 21.72 21.73 - 25.38 25.39 - 29.27 29.28 - 32.34 32.35 - 35.65

25 Photographer: John van den Hoff, Australian GovernmentDivision, © Commonwealth of Australia Photographer: John van den Hoff, Antarctic

Secondary regionalisation The concentration of satellite-observed sea The Workshop agreed that the surface chlorophyll was explored using a data bioregionalisation should ideally differentiate layer comprising log transformed chlorophyll fi rst between the main divisions of coastal a densities (Figure 16). The chlorophyll Antarctica (shelf and slope areas), sea ice distribution was truncated at 10 mg.m-3 zone and northern open ocean waters before (where all values greater than 10 were made further subdividing according to secondary equal to 10), because the variability in higher features. Nevertheless, two potential order productivity most likely results from components of a secondary classifi cation variability in the range from 0-10 mg.m-3. were explored to determine if there is While chlorophyll a concentration may not suffi cient spatial heterogeneity to warrant refl ect primary production absolutely, it was a further subdivision. considered to be a suitable proxy for the purposes of exploring spatial heterogeneity Sea ice was considered to modify the pelagic in primary production at the large scale. environment both in terms of the potential for primary production as well its infl uence Descriptions of each of these datasets are on the distribution of marine mammals provided in Appendix III. and birds. The impact of sea ice on the environment was explored using a data layer comprising the number of days an area was covered by at least 15% concentration of sea ice (Figure 15).

26 FIGURE 15: Proportion (0-1) of the year for which the ocean is covered by at least 15% sea ice. Calculated from satellite-derived estimates of sea ice concentration spanning 1979–2003. (Comiso, 1999

Proportion (0-1)

0 0.01 - 0.08 0.09 - 0.31 0.32 - 0.49 0.50 - 0.63 0.64 - 0.76 0.77 - 0.91 0.92 - 1

FIGURE 16: Mean summer (DEC-FEB) surface chlorophyll-a concentrations. Summer means (1998-2004) from SeaWiFS.

Chlorophyll-a concentration (mg.m-3)

0.06 0.07 - 0.22 0.23 - 0.37 0.38 - 0.53 0.54 - 0.68 0.69 - 0.99 1.00 - 1.77 1.78 - 39.80

27 Photograph by Nick Gales, Australian GovernmentDivision © Commonwealth of Australia Antarctic

28 3.2 Results of Southern Ocean bioregionalisation

Primary regionalisation the northern open ocean waters. The analysis The results of the primary regionalisation highlights the different environmental are shown in Figures 17 (dendrogram) characteristics of large regions including the and 18 (map). The physical properties of continental shelf and slope, frontal features each region are shown in Table 1. This (Subantarctic Front, Polar Front, Southern regionalisation clearly differentiates, at the Antarctic Circumpolar Current Front), the highest levels, between coastal Antarctica deep ocean, banks and basins, island groups, (including embayments), the sea ice zone and and gyre systems.

FIGURE 17: Dendrogram for primary (14-cluster) regionalisation, with thumbnail maps showing regionalisations at different stages of the hierarchy 29 FIGURE 18: Primary regionalisation of the Southern Ocean based on: depth, sea surface temperature (SST), silicate (Si) and nitrate (NOx) concentrations (14 cluster groups) (white areas represent cells with missing data that were not classifi ed in these analyses).

30 TABLE 1: Physical properties (mean and standard deviation of data values) of regions shown in Figure 18 (14 cluster groups based on primary datasets)

REGION NAME Number of Depth Depth SD SST SST SD Si mean Si SD NOx mean NOx SD grid cells mean (m) mean (°C) (µmol/kg) (µmol/kg) Southern Temperate 110567 -4119.952 821.342 8.681 1.854 7.998 2.402 20.919 1.616 Subantarctic Front 40180 -3917.738 921.884 5.840 0.791 15.231 2.582 25.158 1.052 Polar Front 83006 -4134.095 732.582 3.539 0.999 28.382 6.492 29.236 1.815 Southern ACC Front 108053 -4109.261 818.366 0.945 0.872 56.089 9.814 32.370 1.503 Antarctic Open Ocean 136360 -3612.533 897.680 -0.682 0.535 79.593 5.804 33.169 1.374 Antarctic Shelves 30767 -520.048 213.352 -1.149 0.380 82.044 9.211 32.356 1.821 Antarctic Shelf Slope, BANZARE Bank 6508 -1455.466 389.636 -1.227 0.434 79.961 2.946 33.599 1.343 Campbell Plateau, Patagonian Shelf, Africana Rise 7451 -1034.451 427.437 8.453 1.129 7.876 2.582 20.898 1.735 Inner Patagonian Shelf, Campbell & Crozet islands 913 -343.482 109.436 7.742 0.827 8.084 2.233 20.857 1.427 Kerguelen, Heard & McDonald Islands 2294 -1270.202 734.782 3.360 0.818 25.846 4.024 29.279 1.318 Subtropical Front 94234 -4461.472 788.887 11.804 1.511 4.607 1.235 15.257 2.062 Northern Temperate 9946 -4163.621 951.003 15.496 0.774 4.336 0.727 10.154 1.667 Weddell Gyre & Ross Sea banks 52905 -4466.641 762.290 -0.680 0.333 98.163 5.615 31.965 0.553 Chatham Rise 3025 -1568.439 858.953 14.361 0.802 4.112 0.610 12.061 1.453

Uncertainty uncertainty values (red, close to 1) indicate observed sea surface chlorophyll was The time available to the workshop did not that a grid cell lies on the environmental explored using a data layer comprising permit a rigorous analysis of uncertainty. boundary between two different clusters, log transformed chlorophyll densities. and so its allocation to one or the other However, a limited analysis was undertaken The ice and chlorophyll data were is less certain than for a grid cell that is to investigate the uncertainty associated incorporated both separately and in a strongly typical of the cluster to which is has with the clustering algorithm. Figure 19 single classifi cation, and the results of been allocated. Note that this uncertainty illustrates this uncertainty. Uncertainty these analyses are displayed in Appendix IV. analysis considers only a specifi c subset of was computed by fi rst calculating the The preliminary results of this analysis the possible sources of uncertainty in the difference between the environmental using a large number of clusters, based on regionalisation (specifi cally, to do with the characteristics of a grid cell and the average both ice and chlorophyll, are presented in allocation of grid cells to particular clusters). environmental characteristics of the cluster Section 3.3 for three sectors of the Southern to which it was assigned. (Each grid cell Secondary regionalisation Ocean. This exploratory classifi cation is of is assigned to the cluster to which it is The secondary regionalisation incorporated use in illustrating the heterogeneity arising most environmentally similar). A second two additional datasets to refl ect properties from these properties at a smaller scale difference was then computed, this time that further modify the marine environment. than that of the primary regionalisation, between the environmental characteristics The impact of sea ice on the environment however further work is needed to identify of a grid cell and the average environmental was explored using a data layer comprising the appropriate level of regional separation characteristics of the next-most similar the proportion of the year (0-1) that an area using these secondary datasets, and to cluster. The fi rst difference value was then was covered by at least 15% concentration determine whether other datasets could be divided by the second. Thus, high of sea ice. The concentration of satellite- used to assist this process.

Cluster membership uncertainity (proportion)

0.01 - 0.22 0.23 - 0.33 0.34 - 0.43 0.44 - 0.52 0.53 - 0.63 0.64 - 0.74 0.75 - 0.87 0.88 - 1.00

FIGURE 19: Map showing scaled uncertainty for the primary (14-cluster) classifi cation 31 3.3 Expert review of bioregionalisation results

An assessment of the fi nal results was currents, eddies and mixing associated island chains, probably result from the carried out by expert review to determine with the ACC and the Weddell-Scotia mixing of micronutrients with surface waters if the defi ned regions were consistent with Confl uence (WSC) occur in the vicinity through the fl ow of the ACC and the WSC present knowledge of the ecosystem. The of the Scotia Arc. In the central and eastern as they pass over the Scotia Arc. A range of following sections describe the defi ned areas, there is a greater contribution of the zooplankton species including Antarctic krill regions in further detail, focusing separately Weddell gyre and a broadening of the ACC. (Euphausia superba), consume this primary on the Atlantic, Indian and Pacifi c ocean A large continental shelf area is present in production. In turn, these taxa are consumed sectors (CCAMLR Statistical Areas 48, 58 the great embayment of the Weddell Sea, by numerous species of nekton, seabirds and and 88, respectively). For each sector, a along with a number of ice shelves. These marine mammals. The resulting biodiversity map of the regions defi ned by the primary features are captured well in the primary is possibly higher than elsewhere in the regionalisation is overlain with information regionalisation. The Atlantic sector is Southern Ocean. on known large-scale physical and also dominated by strong seasonal cycles, Zooplankton community structure in ecological features such as fronts, gyres, manifest by changing irradiance and the southwest Atlantic appears to be seamounts and maximum sea ice extent. seasonal sea-ice cover. The bathymetry of dependent upon the timing of the seasonal In addition, maps showing an example of the southwest Atlantic steers the fl ow of sea-ice retreat (Ward et al., 2003). Sea-ice a secondary regionalisation (using ice and the ACC northwards, carrying polar waters infl uences the timing of reproduction; chlorophyll data to defi ne 40 clusters for the to more northerly latitudes than elsewhere a late retreat delays reproduction and Southern Ocean) illustrate the high degree in the Southern Ocean. This transport is reduces population sizes of a number of smaller-scale heterogeneity arising critical to the local marine systems around of zooplankton species. During years from patchiness in chlorophyll and sea ice some of the more northerly SubAntarctic of normal sea-ice retreat copepods are concentrations, particularly in shelf areas island groups where large colonies of many more advanced, and there are also higher and the seasonal ice zone. land-based predators breed. abundances of krill larvae. This implies South Atlantic (Area 48) The southwest Atlantic is possibly the most that the seasonal environment critically The Atlantic sector is characterised by the studied of all the areas in the Southern infl uences the biogeography of zooplankton narrowing of the ACC as it passes through Ocean. It has higher productivity than other communities in the southwest Atlantic. Drake Passage between South America and areas. Extensive summer phytoplankton As a consequence, multi-year datasets that the Antarctic Peninsula. In the west, strong blooms, particularly around some of the encompass years of differing environmental

FIGURE 20: Map showing primary regionalisation for the South Atlantic sector (Area 48), with major physical features

32 conditions are likely to provide broader and southwest Atlantic. Small copepods form ice are thus key factors in the high krill more generic descriptions of community approximately 75% of total copepod densities observed in the southwest Atlantic structure in the southwest Atlantic than do abundance in the upper ocean layers across Ocean. Salps, by contrast, occupy the single year synoptic surveys. all major oceanographic zones. These extensive lower-productivity regions of the species show a continuum of temperature Southern Ocean and tolerate warmer waters A number of such analyses are now ranges, and there is no evidence that the than krill (Atkinson et al., 2004). available, which reveal that (multi-year) Polar Front is a major biogeographic sampling of zooplankton species across The secondary regionalisation shows the boundary to their distribution. Indeed, the ACC to the northwest of South Georgia patchiness in primary production and ice several important species reach maximum form four community groupings, and that cover around the coastal region as well as numbers in this area. Total copepod these are geographically consistent with the patchiness in primary production in abundance is thus higher in the vicinity the different water masses identifi ed on the the oceanic areas and around the islands of the Polar Front than in any other region basis of temperature and salinity properties. of the Scotia Arc. The pattern of clusters (Atkinson and Sinclair, 2000). Copepods are the largest contributors to from Antarctic Peninsula area to South total abundance within these groupings. Larger zooplankton species such as Sandwich Island area matches well with All groups can be characterised by varying Antarctic krill and salps (mainly Salpa the spatial distribution of krill length proportions of a relatively small subset thompsoni) are also major grazers in the composition cluster groups observed of species, many of which are present Southern Ocean and particularly in the during CCAMLR-2000 survey (Figure 5 throughout the region. Other species are productive southwest Atlantic sector where of Sigel et al. 2004). Also, the infl uence of characteristic of particular groups. The krill biomass forms more than 50% of the Weddell Gyre on the productivity and close physical and biological coupling Southern Ocean krill stocks. Spatially, water properties of the Bransfi eld Strait observed across the ACC confi rms that summer krill density correlates positively (Amos, 2001) is apparent. frontal zones, and particularly the Polar with chlorophyll a concentrations. Figures 20 and 21 show the primary and Front, are features across which community Temporally, within the southwest Atlantic, secondary regionalisations for the South properties change in the Atlantic sector. summer krill densities correlate positively Atlantic sector. (Note that the colours used with sea-ice extent the previous winter. Small and mesopelagic zooplankton in the secondary regionalisation do not relate Summer food and the extent of winter sea species also play a major role in the to those in the primary regionalisation).

FIGURE 21: Map showing secondary regionalisation for the South Atlantic sector (Area 48)

33 FIGURE 22: Map showing primary regionalisation for the Indian Ocean sector (Area 58), with major physical features

Indian Ocean (Area 58) of 60°E, although a recent survey of shelf is dominated by the crystal krill The Indian Ocean sector extends in the krill and associated environmental Euphausia crystallorophias, which are west from the eastern margins of the parameters (Jan-March 2006 - 30°E to never found as adults to the Weddell Gyre across the Indian Ocean 80°E) completed synoptic coverage of north of the shelf break. the coastal region from 30°E to 150°E with a gradual movement of the Polar A number of studies have characterised the (Nicol et al., 2000; 2006). These surveys Front from the north to the south. The zooplankton assemblages in the Southern suggest that productivity is higher and, fl ow of the ACC is disrupted by the greater Indian Ocean and their association with along with Antarctic krill, Euphausia Kergeulen Plateau, including BANZARE fronts (e.g. Hosie 1994a; 1994b; Hosie superba, extends further to the north in the Bank, causing formation of many branches et al., 1997; Chiba et al. 1999; 2001; area to the west of approximately 115°E of the fronts. A set of subantarctic islands Hoddell et al., 2000; Hosie et al., 2000; (south of 60°S) compared to east of this and banks are found in the western part of Hunt and Hosie, 2003; 2005; 2006a; o o longitude. Salps are found to the north of the sector between 45 S and 55 S. Gyres 2006b). Continuous Plankton recorder this krill distribution. This is evident in are present close to the small embayment (CPR) monitoring, primarily between the secondary classifi cation with a greater of Prydz Bay, and also east of BANZARE 60 and 160°E from spring to autumn, diversity of regions in the area of higher Bank. These general features are evident have identifi ed a number of breaks in the productivity between 115°E and Prydz in the primary classifi cation. distribution of zooplankton taxa with fronts Bay. This is also coincident with the Ice shelves are present along the coastal identifi ed by Orsi et al. (1995). The SAF evidence for an eastern gyre hypothesised margins, including in Prydz Bay. Also, is a major biogeographic boundary for by Nicol et al. (2000). The pattern of a number of polynyas occur along the plankton with separate communities north clusters around 30-80°E (south of 60°S) coast (Arrigo and van Dijken, 2003; and south of the front (Hunt and Hosie, matches well the spatial pattern of krill Massom et al., 1998), some of which are 2003). Some changes in composition length composition cluster groups found substantial contributors to production of occur at the Polar Front, in particular the in recent surveys (Figure 12 of Nicol et sea ice and deep water formation. The northern branch of the PF. A number of al., 2006). These surveys also identifi ed annual progression and retreat of sea ice is distinct zooplankton assemblages can the higher densities of Antarctic krill, uninterrupted in this region, extending to also be defi ned south of the PF-N. These Euphausia superba, associated with the 60oS, although the winter sea ice extent is assemblages are identifi ed more by subtle shelf break. This region also is indicated greater in the west than in the east. variation in the abundance or proportion well in the secondary classifi cation. The of species rather than changes in species Most research has occurred to the west neritic community over the continental 34 FIGURE 23: Map showing secondary regionalisation for the Indian Ocean sector (Area 58)

composition (Hunt and Hosie, 2005). The Polar Fontal Zone (SAF and PF) is often reported as an area of elevated primary production which then declines south of the PF. This is probably true for phytoplankton, and certainly many vertebrate predators forage in this area. However, the CPR survey has consistently shown that zooplankton abundance increases substantially south across the SAF and remain high though the Southern Ocean to a point between 60 to 62°S where zooplankton abundance declines suddenly (Hosie et al., 2003). The upper 20 m of the water column in the area further south usually remains almost devoid of zooplankton. This decline approximates the position of the SACCF (Orsi et al., 1995) although a link is yet to be established. Overall, the patterns displayed in the secondary classifi cation correspond to the patterns of zooplankton described here.

Figures 22 and 23 show the primary and secondary regionalisations for the Indian Ocean sector. (Note that the colours used in the secondary regionalisation do not relate to those in the primary regionalisation).

35 FIGURE 24: Map showing primary regionalisation for the Pacifi c Ocean sector (Area 88), with major physical features

Pacifi c Ocean (Area 88) is dominated by Phaeocystis (Ainley et al., Campbell Plateau is not differentiated from The Pacifi c sector is similar in ocean 2006). The primary classifi cation identifi es the southern temperate waters. This may be characteristics to the Indian Ocean sector the Ross Sea shelf and slope areas. Features because of the lack of differentiation in the except for the interaction with Ross Sea such as the Ross Sea polynya are not chlorophyll data used here. captured in the primary classifi cation, but and its associated gyre. The inner Ross Sea The secondary classifi cation does identify these may be refl ected in the heterogeneity over the continental shelf has characteristics the heterogeneity of the environment of the secondary classifi cation. distinct from those of the ACC. The western associated with the island and ridge system part of the Ross Sea has a complex shelf and Further to the east in the Pacifi c Sector in the eastern part of the sector. It also slope area along with the Balleny Islands is a narrowing of the ACC towards the identifi es the expected complexity in the and ridges of seamounts extending to the Drake Passage. Also, the seasonal sea Ross Sea Gyre and its relationship to the north (the Macquarie Ridge extending to ice zone narrows in the eastern part of coastal system. These results refl ect studies the Campbell Plateau) and to the east. A the Bellingshausen Sea. The primary documenting the variation in diversity and clockwise current fl ows within the area classifi cation captures the major ocean and ecological processes in the region (Bradford- shallower than the 500 m isobath, and the coastal features, although it does not refl ect Grieve and Fenwick 2001; Pinkerton et East Wind Drift current fl ows in the opposite the ocean ridges in the eastern part of al., 2006; Sharp, 2006). For example, direction along the continental shelf break. the sector. Euphuasia superba is found to the north Upwelling of Circumpolar Deep Water also of the shelf break while the neritic fauna is The separation of subtropical and occurs along this shelf break (Ainley, 2002). dominated by Euphausia crystallorphias and subantarctic waters as well as distinguishing Seasonal polynyas in the western shelf area Pleuragramma antarcticum. play an important role in the distribution of the Campbell Plateau from the ocean phytoplankton, zooplankton, fi sh, birds and environment is supported by research on Figures 24 and 25 show the primary and seals. The concentration of top predators in productivity of the region (Boyd et al. 1999; secondary regionalisations for the Pacifi c the Ross Sea coincides with the marginal Murphy et al. 2001) and fi sh assemblages Ocean sector. (Note that the colours used in ice zone that rings the Ross Sea Polynya. (Bradford-Grieve et al. 2003). Although the secondary regionalisation do not relate to This area is dominated by diatoms, while the these general differences are retained in those in the primary regionalisation). central, open water portion of the polynya the secondary classifi cation, the wider

36 FIGURE 25: Map showing secondary regionalisation for the Pacifi c Ocean sector (Area 58)

The primary regionalisation of the Southern Ocean highlights the different environmental characteristics of large regions including the continental shelf and slope, frontal features (Subantarctic Front, Polar Front, Southern Antarctic Circumpolar Current Front), the deep ocean, banks and basins, island groups and gyre systems. The addition of secondary datasets suggests smaller-scale spatial heterogeneity within the regions, particularly in the continental shelf and slope areas, and the seasonal ice zone.

37 Lyn Irvine, Australian GovernmentDivision, © Commonwealth of Australia Antarctic Lyn

38 4. Future work

The workshop identifi ed a range of areas other datasets could be used to assist this be used in defi ning the classifi cation. in which future work might be directed in process. In particular, sea ice is a major It may be important to consider stochastic, order to produce a fi nal bioregionalisation driver of ecosystem processes in the Southern temporal or spatial variability in defi ning for the Southern Ocean. Priorities included Ocean (see Section 1.3), and the inclusion of bioregions in order to ensure that the outcome the incorporation of additional (particularly variables representing sea ice dynamics is robust to uncertainties and variability. biological) datasets, and fi ner-scale analysis will be important in fi ner-scale Further work towards understanding (and, of particular areas of interest. It also identifi ed regionalisation analyses. where possible, reducing) different types of that the statistical methods might be refi ned The draft regionalisation presented in uncertainty in the data, models or methods further. However, the workshop was satisfi ed this report is pelagic, however it may will help the classifi cation process. that the proof of concept developed is also be necessary to undertake a benthic suffi cient to undertake the tasks identifi ed A bioregionalisation will inevitably be based regionalisation. Further consideration should by CCAMLR and the CEP. on the best scientifi c evidence available at the be given to the relationships between the time. Further refi nement could be achieved Further data could be used to update or refi ne benthic and pelagic systems, and the utility by adding biological and environmental data the draft broad-scale primary regionalisation of separating the two systems in the context as it becomes available, thereby reducing (for example using additional datasets such of bioregionalisation analysis. uncertainties. One source of refi nement as insolation (PAR) and sea surface height). A range of potential biological datasets (for will be to add more biological data to test Datasets used in this analysis might also be use in future analyses) were identifi ed during the relationship between physical and refi ned, for example using derived datasets the workshop (see Appendix V), but it will be environmental surrogates and the ecological such as a ‘silicate depletion’ data layer necessary to identify which of these would processes they are thought to represent. (refl ective of primary production) derived by be of most value. Data ‘compendia’ may be This is likely to be needed at fi ner-scale subtracting silicate at the surface from silicate of assistance in providing information on resolution of the bioregionalisation. at 200 m depth.The remotely-sensed PAR data inter-annual and seasonal variability, which are confounded by the inability to distinguish The most important avenue for further can then be further analysed according to ice cover from cloud cover, however they work will be to undertake a fi ner-scale the defi ned objectives. Indicator species might be transformed to represent biologically regionalisation than that presented here. might also be investigated for their potential relevant variation in available light, by This might initially be focused in areas utility in providing further input to the combining PAR data and ice data. where more data is available, such as in the analysis. In the longer term, the compilation southwest Atlantic. The addition of datasets on The addition of sea ice and chlorophyll of comprehensive biological data sets may chlorophyll a and sea ice extent illustrated the a datasets in the exploratory secondary allow the use of more sophisticated analytical complexity of the coastal, shelf and seasonal classifi cations illustrated the high level of approaches such as Generalised Dissimilarity ice areas, in relation to these parameters. heterogeneity arising from these parameters. Modelling (GDM – Ferrier et al. in press). These regions are likely to have additional Refi nement of the analysis and data used at This performs an integrated statistical complexity corresponding to other ecological the secondary classifi cation level is needed analysis of biological and environmental processes and species distributions, and to identify the appropriate level of regional data, using information on species turnover should be a priority for further research. separation at a smaller scale using these rates to identify the optimal weighting and secondary datasets, and to identify whether transformation of environmental variables to

5. Conclusions This workshop established a ‘proof of concept’ for bioregionalisation of the Southern Ocean. Further work within the frameworks of CCAMLR and the CEP will be an important contribution to the achievement of a range of scientifi c, management and conservation objectives, including large-scale ecological modelling, ecosystem-based management of human activities in the marine environment, and the development of ecologically representative protected area systems. Continuing work on this topic also has the potential to inform and contribute to the further development of bioregionalisation analysis as a tool for conservation and management in the global context.

39 List of Appendices (provided on CD)

APPENDIX I: Approaches to bioregionalisation – examples presented during the workshop

• Antarctic Environmental Domains Analysis (Harry Keys and Fraser Morgan, Department of Conservation, New Zealand)

• CCAMLR Small-Scale Management Units for the fi shery on Antarctic krill in the Southwest Atlantic (Roger Hewitt, NOAA, USA)

• Australian National Bioregionalisation: Pelagic Regionalisation (Vincent Lyne and Donna Hayes, Department of the Environment and Heritage and CSIRO)

• Selecting Marine Protected Areas in New Zealand’s EEZ (John Leathwick, NIWA, New Zealand)

APPENDIX II: Technical information on approach to bioregionalisation

APPENDIX III: Descriptions of datasets used in the analysis

APPENDIX IV: Results of secondary regionalisation using ice and chlorophyll data

APPENDIX V: Biological datasets of potential use in further bioregionalisation work

APPENDIX VI: Details of datasets, Matlab code and ArcGIS shapefi les included on the CD

40 List of workshop participants

PARTICIPANT EXPERTISE AFFILIATION Dr. Ian Ball Mathematical modelling, ACE CRC, MPA selection AGAD, Australia Ms. Kendall Benton Support WWF-Australia Dr. Andrew Constable Chair, ecosystem ecology ACE CRC, and modelling, marine AGAD, Australia protected areas, CCAMLR Dr. Susan Doust Support, CCAMLR AGAD, Australia Dr. Susie Grant Support, marine protected UK areas, CCAMLR, CEP Dr. Roger Hewitt Krill biology, ecosystem SW Fisheries Centre, ecology NOAA, USA Dr. Graham Hosie Zooplankton ecology; AGAD, Australia CPR program SCAR representative Dr. So Kawaguchi Krill biology, ACE CRC, AGAD, ecosystem ecology Australia Dr. Harry Keys Bioregionalisation, Antarctic DOC, NZ regional classifi cation Dr. John Leathwick Statistical modelling, NIWA, bioregionalisation, MPA selection Hamilton, NZ Dr. Gilly Llewellyn Support, classifi cation, WWF-Australia regionalisation, marine protected areas Dr. Vincent Lynne Statistician, bioregionalisation CSIRO MAR, Hobart, Australia Dr. Enrique Marschoff Krill biologist (larvae), CCAMLR Argentina Dr. Rob Massom Sea ice, remote sensing ACE CRC, Australia Mr. Ewan McIvor Environmental policy AGAD, Australia Dr. Denzil Miller Observer CCAMLR Secretariat Dr. Mikio Naganobu Krill biology, ecosystem ecology Japan Dr. Steve Nicol Krill biology, ecosystem ecology ACE CRC, AGAD, Australia Dr. Matt Pinkerton Ecosystem modelling, NIWA, NZ remote sensing Dr. Ben Raymond Modelling, data visualisation ACE CRC, AGAD, Australia Dr. Steve Rintoul Physical oceanography, ACE CRC, Australia remote sensing Dr. Eugene Sabourenkov Observer CCAMLR Secretariat Dr. Ben Sharp Marine biologist, marine MFish, NZ protected area research Dr. Sergei Sokolov Physical oceanography, ACE CRC, Australia remote sensing Mr. Rick Smith Statistician, bioregionalisation CSIRO MAR, Hobart, Australia Dr. Pete Strutton Primary production, Oregon State remote sensing University, USA Dr. Phil Trathan Ecosystem and predator BAS, UK ecology, CCAMLR

41 Glossary of terms

Bioregionalisation (or Regionalisation) Parameter A process that aims to partition a broad Information extracted from data. For spatial area into distinct spatial regions, example, sea ice concentration is a variable using a range of environmental and from which the parameters of ‘proportion biological information. The process of year when the ocean is covered by at results in a set of bioregions, each with least 15% ice’ or ‘areas with greater than relatively homogeneous and predictable 50% ice coverage’ can be extracted. ecosystem properties. The properties of Property a given bioregion should differ from those of adjacent regions in terms of species This term is used here to describe the composition as well as the attributes defi ning characteristics or attributes of of its physical and ecological habitats. a particular ecological process, or of a The term regionalisation may be used given region. interchangeably (or sometimes to refer Proxy to an analysis undertaken using only physical data). A parameter that can be used to provide similar information or patterns to another Bioregion (or Region) parameter or variable, usually used when A spatial compartment defi ned on the basis desired data (e.g. the distribution of species) of its biological and/or physical properties. are unavailable, or where one parameter Each bioregion (or region) refl ects a unifying can be used in the place of several others set of major environmental infl uences in order to simplify the analysis. which shape the occurrence of biota Site and their interaction with the physical environment. The term region may be used In the context of bioregionalisation analysis, interchangeably (or sometimes to refer to a site is the smallest unit of analysis. The spatial compartments which have been study area is divided into a grid of sites, at defi ned using only physical data). suffi ciently fi ne scale to enable appropriate resolution of the fi nal areas. Each site Classifi cation will have a particular set of parameters, The process of partitioning a broad spatial according to the input data. area into distinct regions. Also used to refer Synoptic to the specifi c step within that process during which the actual allocation of This term is used to describe data that have sites to regions occurs, usually through a broad and continuous spatial coverage (e.g. statistical process such as cluster analysis across the entire Southern Ocean). Here, synoptic data may also refer to summaries Ecological process of the observed conditions over time (e.g. In the context of this report, an ecological mean monthly values averaged to obtain process is any process that affects the an annual mean). Synoptic data may be dynamics of a species obtained through satellite remote-sensing, or through model climatologies generated Hierarchy (spatial and statistical) from observed values. In the context of bioregionalisation, this Uncertainty term may be used to refer to spatial or ecological hierarchy, or statistical hierarchy. In the context of bioregionalisation analysis, uncertainty refers to the effects Spatial or ecological hierarchy refers to of imprecision in data (e.g. measurement the different levels of scale or ecological error), uncertainty within models used to processes within a large area. A hierarchy derive variables (e.g. climatology models), may be nested, whereby smaller scale and epistemic uncertainty (potential errors units or processes are nested within large in the chosen method). Each of these scale units. types of uncertainty can affect the resulting Statistical hierarchy has relevance in region boundaries. dissimilarity-based clustering methods, Variable where an iterative approach is undertaken to group sites together into regions. All Variables are physical or environmental sites are initially allocated to their own data from which specifi c parameters regions. At each iteration of the process can be extracted. For example, sea ice (or each step down the hierarchy), the two concentration is a variable from which the most similar regions are merged together, parameters of ‘proportion of year when the until at the end of the process there is only ocean is covered by at least 15% ice’ or one region, which contains all of the sites. ‘areas with greater than 50% ice coverage’ This is often displayed in dendograms. can be extracted.

42 References

Ackley, S.F., and Sullivan, C.W. (1994) Physical controls on the development and characteristics of Antarctic sea ice biological communities a review and synthesis. Deep-Sea Research 41(10): 1583 1604

Ainley, D.G. (2002) The Ross Sea, Antarctica, where all ecosystem processes still remain for study, but maybe not for long. Marine Ornithology 30: 55-62

Ainley, D. G., Tynan, C.T. and Stirling, I. (2003) Sea ice: A critical habitat for polar marine mammals and birds. In D. Thomas and G. Dieckmann (eds), Sea Ice – An Introduction to its Physics, Biology, Chemistry and Geology. Blackwell Science, Oxford, pp. 240-266.

Ainley, D., Toniolo, V., Ballard, G., Barton, K., Eastman, J., Karl, B., Focardi, S., Kooyman, G., Lyver, P., Olmastroni, S., Stewart, B.S., Testa J.W. and Wilson, P. (2006) Managing ecosystem uncertainty: critical habitat and dietary overlap of top-predators in the Ross Sea. WG-EMM-06/29.

Amos, F.A. (2001) A decade of oceanographic variability in summertime near Elephant Island, Antarctica. Journal of Geophysical Research 106 (C10): 22401–22423.

Andrewartha, H.G. and Birch, L. C. (1984) The ecological web: more on the distribution and abundance of animals. University of Chicago Press, Chicago.

Arrigo, K.R., and van Dijken, G.L. (2003) Phytoplankton dynamics within 37 Antarctic coastal polynya systems. J. Geophys. Res. 108(C8): 3271. doi:10.1029/2002JC001739.

ATCM XXIV/WP012 (2001) Systematic Environmental-Geographic Framework for Protected Areas under Annex V of the Environmental Protocol. Antarctic Treaty Consultative Meeting XXIV. Working Paper No. 12.

Atkinson, A. and Sinclair, J.D. (2000) Zonal distribution and seasonal vertical migration of copepod assemblages in the Scotia Sea. Polar Biology 23(1): 46-58

Atkinson, A., Siegel, V., Pakhomov, E. and Rothery, P. (2004) Long-term decline in krill stock and increase in salps within the Southern Ocean. Nature 432 (7013): 100-103

Bailey, R.G. (1996) Ecosystem Geography. Springer-Verlag, New York.

Beaman R.J. and Harris P.T. (2005) Bioregionalization of the George V Shelf, East Antarctica. Continental Shelf Research 25: 1657–1691

Belbin, L. (1993) Environmental representativeness: regional partitioning and reserve selection. Biological Conservation 66: 223–230.

Bindoff, N. L., Rintoul, S.R. and Massom R. (2000). Polynyas and bottom water formation south of Tasmania. Papers and Proceedings of the Royal Society of Tasmania, 133(3), 51–56

Boyd, P.W., LaRoche, J., Gall, M., Frew, R. and McKay, R. (1999) Role of iron, light, and silicate in controlling algal biomass in subantarctic waters SE of New Zealand. Journal of Geophysical Research 104, C6, pp 13391-13404.

Bradford-Grieve, J. and Fenwick, G. (2001) A review of the current knowledge describing the biodiversity of the Ross Sea. Final research report for Ministry of Fisheries, Research Project ZBD2000/01 Objective 1 (in part). 93 pp.

Bradford-Grieve, J.M., Probert, P.K., Nodder, S.D., Thompson, D., Hall, J., Hanchet, S., Boyd, P., Zeldis, J., Baker, A.N., Best, H.A., Broekhuizen, N., Childerhouse, S., Clark, M., Hadfi eld, M., Safi , K. and Wilkinson, I. (2003) Pilot trophic model for subantarctic water over the Southern Plateau, New Zealand: a low biomass, high transfer effi ciency system. Journal of Experimental Marine Biology and Ecology 289: 223-262.

Brierley, A.S., and Thomas, D.N. (2002) Ecology of Southern Ocean pack ice. Advances in Marine Ecology 43: 171-276

Casey, K.S. and Cornillon, P. (1999) A comparison of satellite and in situ based sea surface temperature climatologies, J. Climate, 12 (6): 1848-1863. http://podaac-www.jpl.nasa.gov/products/product112.html

CEP IX (2006) Report of the Committee for Environmental Protection (CEP IX). Edinburgh, 2006.

Chiba, S., Ishimaru, T., Hosie, G.W., Wright, S.W. (1999) Population structure change of Salpa thompsoni from austral mid-summer to autumn. Polar Biology 22: 341-349

Chiba, S., Ishimaru, T., Hosie, G.W., Fukuchi, M. (2001) Spatio-temporal variability of zooplankton community structure in the western Pacifi c Ocean sector, Antarctica. Marine Ecology Progress Series 216: 95-108.

Clarke, A. (1996) Distribution of Antarctic marine benthic communities. In: Ross, R.M., Hofmann, E.E. & Quetin, L.B. Foundations for Ecological Research West of the Antarctic Peninsula. American Geophysical Union, Washington. (pp. 219-230).

Clarke, A. and Johnston, N.M. (2003) Antarctic Marine Benthic Diversity. Oceanography and Marine Biology: an Annual Review 41: 47-114.

Comiso, J. (1999), updated 2005. Bootstrap sea ice concentrations for NIMBUS-7 SMMR and DMSP SSM/I. Boulder, CO, USA: National Snow and Ice Data Center. Digital media. http://nsidc.org/data/nsidc-0079.html

Constable, A.J. and Nicol, S. (2002) Defi ning smaller-scale management units to further develop the ecosystem approach in managing large-scale pelagic krill fi sheries in Antarctica. CCAMLR Science 9: 117–131.

Dell, R.K. (1972) Antarctic benthos. Advances in Marine Biology 10: 1-216.

Eicken, H. (1992) The role of sea ice in structuring Antarctic ecosystems. Polar Biology 12: 3-13

Ekman, V.K. (1953) Zoogeography of the Sea. Sidgwick and Jackson, London.

Everson, I. (1977) The living resources of the Southern Ocean. Food and Agriculture Organisation, GLO/S0/77/1. Rome.

Ferrier, S., Manion, G., Elith, J., and Richardson, K. in press. Using generalised dissimilarity modelling to analyse and predict patterns of beta-diversity in regional biodiversity assessment. Diversity and Distributions.

Faith, D.P., Minchin, P.R., and Belbin, L. (1987) Compositional dissimilarity as a robust measure of ecological distance. Vegetation 69:57-68.

Fowler, C. (2003) Polar Pathfi nder Daily 25 km EASE-Grid Sea Ice Motion Vectors. US National Snow and Ice Data Center, Boulder, Colorado, USA. Digital media.

Frouin, R., Franz, B. and Wang, M. Algorithm to estimate PAR from SeaWiFS data Version 1.2 – Documentation. http://oceancolor. gsfc.nasa.gov/DOCS/seawifs_par_algorithm.pdf

Garrison, D.L. and Mathot, S. (1996) Pelagic and sea ice microbial communities. In Foundations for Ecological Research West of the Antarctic Peninsula, Ross, R.M., Hofmann, E.E. and Quetin, L.B. (Eds) American Geophysical Union, Antarctic Research Series, 70, pp. 155-172

43 References

Gloersen, P., Campbell, W.J., Cavalieri, D.J., Comiso, J.G., Parkinson, C.L. and Zwally, H.J. (1992). Arctic and Antarctic sea ice, 1978 1987: Satellite passive microwave observa¬tions and analysis. NASA Special Publication SP 511, NASA, Washington D.C., USA, 290 pp.

Gouretski, V.V. and Koltermann, K. P. (2004) WOCE Global Hydrographic Climatology. Technical Report, 35, Berichte des Bundesamtes für Seeschifffahrt und Hydrographie. http://odv.awi-bremerhaven.de/data/ocean/woce-global-hydrographic-climatology.html

Hayes, D., Lyne, V., Condie, S., Griffi ths, B., Pigot, S. and Hallegraef, G. (2005) Collation and Analysis of Oceanographic Datasets for National Marine Bioregionalisation. CSIRO Report to the National Oceans Offi ce. CSIRO Marine Research, Hobart.

Hedgpeth, J.W. (1970) Marine biogeography of the Antarctic regions. In: Holdgate, M.W. (ed.) Antarctic Ecology (Volume 1) Academic Press, London. (pp. 97-104)

Hewitt, R.P., Watkins, J.L., Naganobu, M., Tshernyshkov, P., Brierley, A.S., Demer, D.A., Kasatkina, S., Takao, Y., Goss, C., Malyshko, A., Brandon, M.A., Kawaguchi, S., Siegel, V., Trathan, P.N., Emery, J.H., Everson, I. and Miller, D.G.M. (2002) Setting a precautionary catch limit for Antarctic krill. Oceanography 15(3): 26–33.

Hewitt, R.P., Watters, G., Trathan, P.N., Croxall, J.P., Goebel, M.E., Reid, K., Trivelpiece, W.Z. and Watkins, J.L. (2004). Options for allocating the precautionary catch limit of krill among Small-Scale Management Units in the Scotia Sea. CCAMLR Science 11: 81-98.

Hoddell, R.J., Crossley, A.C., Williams, R. and Hosie, G.W. (2000) The Distribution of Antarctic Pelagic Fish and Larvae (CCAMLR Division 58.4.1) Deep Sea Research 47(12-13): 2519-2541

Hosie, G.W. (1991) Distribution and abundance of euphausiid larvae in the Prydz Bay region, Antarctica (January 1985). Antarctic Science 3(2):167-180

Hosie, G.W. (1994a) The macrozooplankton communities in the Prydz Bay region, Antarctica. In: S.Z. El-Sayed (ed) Southern Ocean Ecology: The BIOMASS Perspective. Cambridge University Press pp 93-123.

Hosie, G.W. (1994b) Multivariate analyses of the macrozooplankton community and euphausiid larval ecology in the Prydz Bay region, Antarctica. ANARE Reports 137.

Hosie, G.W., Cochran, T.G., Pauly, T., Beaumont, K.L., Wright, S.W., Kitchener, J.A. (1997) The zooplankton community structure of Prydz Bay, January-February 1993. Proceedings of the NIPR Symposium on Polar Biology 10: 90-133.

Hosie, G.W., Schultz, M.B., Kitchener, J.A.,Cochran, T.G. and Richards, K. (2000) Zooplankton community structure off East Antarctica (80-150° east) during the Austral summer of 1995/96. Deep Sea Research 47(12-13): 2437-2463

Hosie, G.W., Fukuchi, M. and Kawaguchi, S. (2003) Development of the Southern Ocean Continuous Plankton Recorder Survey. Progress in Oceanography 58(2-4): 263-283

Hunt, B.P.V. and Hosie, G.W. (2003) The Continuous Plankton Recorder in the Southern Ocean: a comparative analysis of zooplankton communities sampled by the CPR and vertical net hauls along 140°E. Journal of Plankton Research 25(12): 1-19

Hunt, B.P.V. and Hosie, G.W. (2005) Zonal Structure of Zooplankton Communities in the Southern Ocean south of Australia: results from a 2150 kilometre Continuous Plankton Recorder transect. Deep-Sea Research I 52: 1241-1271.

Hunt, B.P.V. and Hosie, G.W. (2006a) Seasonal zooplankton community succession in the Southern Ocean south of Australia, Part I: The Seasonal Ice Zone. Deep-Sea Research I 53: 1182-1202

Hunt, B.P.V. and Hosie, G.W. (2006b) Seasonal zooplankton community succession in the Southern Ocean south of Australia, Part II: The Sub-Antarctic to Polar Frontal Zones. Deep-Sea Research I 53: 1203-1223

Ichii, T. (1990) Distribution of Antarctic krill concentrations exploited by Japanese krill trawlers and minke whales. Proceedings of the NIPR Symposium on Polar Biology 3: 36-56

IOC, IHO and BODC, 2003. Centenary Edition of the GEBCO Digital Atlas, published on CD-ROM on behalf of the Intergovernmental Oceanographic Commission and the International Hydrographic Organization as part of the General Bathymetric Chart of the Oceans, British Oceanographic Data Centre, Liverpool, U.K.

Kaufman, L. and Rousseeuw, P.J. Finding Groups in Data. An Introduction to Cluster Analysis. John Wiley & Sons, New York, 1990.

Knox, G.A. (1994) The Biology of the Southern Ocean. Cambridge, Cambridge University Press, 444 p.

Kock, K.-H. (2000) Understanding CCAMLR’s approach to management. CCAMLR, Hobart, Australia.

Kwok, R. and Comiso, J.C. (2002) Southern Ocean climate and sea ice anomalies associated with the Southern Oscillation. Journal of Climate 15(5): 487-501.

Leathwick, J.R., Dey, K. and Julian, K. (2006a) Development of an environmental classifi cation optimised for demersal fi sh. NIWA Client Report HAM2006-063.

Leathwick, J.R., Francis, M.P. and Julian, K. (2006b) Development of a demersal fi sh community map of New Zealand’s Exclusive Economic Zone. NIWA Client Report HAM2006-062.

Leathwick, J.R., Julian, K. and Francis, M.P. (2006c) Exploration of the use of reserve planning software to identify potential Marine Protected Areas in New Zealand’s Exclusive Economic Zone. NIWA Client Report HAM2006-064.

Leathwick, J.R., Overton, J. McC. and McLeod, M. (2003) An environmental domain classifi cation of New Zealand and its use as a tool for biodiversity management. Conservation Biology 17: 1612–1623

Legendre, L., Ackley, S.F., Dieckmann, G.S., Gullicksen, B., Homer, R., Hoshiai, T., Melnikov, I.A., Reeburgh, W.S., Spindler, M. and Sullivan, C.W. (1992) Ecology of sea ice biota: 2. Global Signifi cance. Polar Biology 12: 429 444

Linse, K., Griffi ths, H.J., Barnes, D.K.A. and Clarke, A. (2006) Biodiversity and biogeography of Antarctic and sub-Antarctic mollusca. Deep Sea Research II 53: 985-1008.

Lizotte, M.P. (2001) The contributions of sea ice algae to Antarctic marine primary production. American Zoologist 41(1): 57-73

Lizotte, M. P. and Arrigo, K.R. (eds) (1998) Antarctic Sea Ice: Biological Processes, Interactions, and Variability. Antarctic Research Series 73, American Geophysical Union, Washington DC.

Longhurst, A.R. (1998) Ecological Geography of the Sea. Academic Press, Florida.

Lyne, V and Hayes, D. (2005) Pelagic Regionalisation: National Marine Bioregionalisation Integration Project. CSIRO Report to the National Oceans Offi ce. CSIRO Marine Research, Hobart. (Summary)

Massom, R. A. (2003) Recent iceberg calving events in the Ninnis Glacier region, East Antarctica, Ant. Sc., 15(2): 303-313

Massom, R. A., Harris, P.T., Michael, K.J. and Potter, M.J. (1998) The distribution and formative processes of latent-heat polynyas in East Antarctica. Ann. Glaciol., 27: 420-426

44 References

Gloersen, P., Campbell, W.J., Cavalieri, D.J., Comiso, J.G., Parkinson, C.L. and Zwally, H.J. (1992). Arctic and Antarctic sea ice, 1978 1987: Satellite passive microwave observa¬tions and analysis. NASA Special Publication SP 511, NASA, Washington D.C., USA, 290 pp.

Gouretski, V.V. and Koltermann, K. P. (2004) WOCE Global Hydrographic Climatology. Technical Report, 35, Berichte des Bundesamtes für Seeschifffahrt und Hydrographie. http://odv.awi-bremerhaven.de/data/ocean/woce-global-hydrographic-climatology.html

Hayes, D., Lyne, V., Condie, S., Griffi ths, B., Pigot, S. and Hallegraef, G. (2005) Collation and Analysis of Oceanographic Datasets for National Marine Bioregionalisation. CSIRO Report to the National Oceans Offi ce. CSIRO Marine Research, Hobart.

Hedgpeth, J.W. (1970) Marine biogeography of the Antarctic regions. In: Holdgate, M.W. (ed.) Antarctic Ecology (Volume 1) Academic Press, London. (pp. 97-104)

Hewitt, R.P., Watkins, J.L., Naganobu, M., Tshernyshkov, P., Brierley, A.S., Demer, D.A., Kasatkina, S., Takao, Y., Goss, C., Malyshko, A., Brandon, M.A., Kawaguchi, S., Siegel, V., Trathan, P.N., Emery, J.H., Everson, I. and Miller, D.G.M. (2002) Setting a precautionary catch limit for Antarctic krill. Oceanography 15(3): 26–33.

Hewitt, R.P., Watters, G., Trathan, P.N., Croxall, J.P., Goebel, M.E., Reid, K., Trivelpiece, W.Z. and Watkins, J.L. (2004). Options for allocating the precautionary catch limit of krill among Small-Scale Management Units in the Scotia Sea. CCAMLR Science 11: 81-98.

Hoddell, R.J., Crossley, A.C., Williams, R. and Hosie, G.W. (2000) The Distribution of Antarctic Pelagic Fish and Larvae (CCAMLR Division 58.4.1) Deep Sea Research 47(12-13): 2519-2541

Hosie, G.W. (1991) Distribution and abundance of euphausiid larvae in the Prydz Bay region, Antarctica (January 1985). Antarctic Science 3(2):167-180

Hosie, G.W. (1994a) The macrozooplankton communities in the Prydz Bay region, Antarctica. In: S.Z. El-Sayed (ed) Southern Ocean Ecology: The BIOMASS Perspective. Cambridge University Press pp 93-123.

Hosie, G.W. (1994b) Multivariate analyses of the macrozooplankton community and euphausiid larval ecology in the Prydz Bay region, Antarctica. ANARE Reports 137.

Hosie, G.W., Cochran, T.G., Pauly, T., Beaumont, K.L., Wright, S.W., Kitchener, J.A. (1997) The zooplankton community structure of Prydz Bay, January-February 1993. Proceedings of the NIPR Symposium on Polar Biology 10: 90-133.

Hosie, G.W., Schultz, M.B., Kitchener, J.A.,Cochran, T.G. and Richards, K. (2000) Zooplankton community structure off East Antarctica (80-150° east) during the Austral summer of 1995/96. Deep Sea Research 47(12-13): 2437-2463

Hosie, G.W., Fukuchi, M. and Kawaguchi, S. (2003) Development of the Southern Ocean Continuous Plankton Recorder Survey. Progress in Oceanography 58(2-4): 263-283

Hunt, B.P.V. and Hosie, G.W. (2003) The Continuous Plankton Recorder in the Southern Ocean: a comparative analysis of zooplankton communities sampled by the CPR and vertical net hauls along 140°E. Journal of Plankton Research 25(12): 1-19

Hunt, B.P.V. and Hosie, G.W. (2005) Zonal Structure of Zooplankton Communities in the Southern Ocean south of Australia: results from a 2150 kilometre Continuous Plankton Recorder transect. Deep-Sea Research I 52: 1241-1271.

Hunt, B.P.V. and Hosie, G.W. (2006a) Seasonal zooplankton community succession in the Southern Ocean south of Australia, Part I: The Seasonal Ice Zone. Deep-Sea Research I 53: 1182-1202

Hunt, B.P.V. and Hosie, G.W. (2006b) Seasonal zooplankton community succession in the Southern Ocean south of Australia, Part II: The Sub-Antarctic to Polar Frontal Zones. Deep-Sea Research I 53: 1203-1223

Ichii, T. (1990) Distribution of Antarctic krill concentrations exploited by Japanese krill trawlers and minke whales. Proceedings of the NIPR Symposium on Polar Biology 3: 36-56

IOC, IHO and BODC, 2003. Centenary Edition of the GEBCO Digital Atlas, published on CD-ROM on behalf of the Intergovernmental Oceanographic Commission and the International Hydrographic Organization as part of the General Bathymetric Chart of the Oceans, British Oceanographic Data Centre, Liverpool, U.K.

Kaufman, L. and Rousseeuw, P.J. Finding Groups in Data. An Introduction to Cluster Analysis. John Wiley & Sons, New York, 1990.

Knox, G.A. (1994) The Biology of the Southern Ocean. Cambridge, Cambridge University Press, 444 p.

Kock, K.-H. (2000) Understanding CCAMLR’s approach to management. CCAMLR, Hobart, Australia.

Kwok, R. and Comiso, J.C. (2002) Southern Ocean climate and sea ice anomalies associated with the Southern Oscillation. Journal of Climate 15(5): 487-501.

Leathwick, J.R., Dey, K. and Julian, K. (2006a) Development of an environmental classifi cation optimised for demersal fi sh. NIWA Client Report HAM2006-063.

Leathwick, J.R., Francis, M.P. and Julian, K. (2006b) Development of a demersal fi sh community map of New Zealand’s Exclusive Economic Zone. NIWA Client Report HAM2006-062.

Leathwick, J.R., Julian, K. and Francis, M.P. (2006c) Exploration of the use of reserve planning software to identify potential Marine Protected Areas in New Zealand’s Exclusive Economic Zone. NIWA Client Report HAM2006-064.

Leathwick, J.R., Overton, J. McC. and McLeod, M. (2003) An environmental domain classifi cation of New Zealand and its use as a tool for biodiversity management. Conservation Biology 17: 1612–1623

Legendre, L., Ackley, S.F., Dieckmann, G.S., Gullicksen, B., Homer, R., Hoshiai, T., Melnikov, I.A., Reeburgh, W.S., Spindler, M. and Sullivan, C.W. (1992) Ecology of sea ice biota: 2. Global Signifi cance. Polar Biology 12: 429 444

Linse, K., Griffi ths, H.J., Barnes, D.K.A. and Clarke, A. (2006) Biodiversity and biogeography of Antarctic and sub-Antarctic mollusca. Deep Sea Research II 53: 985-1008.

Lizotte, M.P. (2001) The contributions of sea ice algae to Antarctic marine primary production. American Zoologist 41(1): 57-73

Lizotte, M. P. and Arrigo, K.R. (eds) (1998) Antarctic Sea Ice: Biological Processes, Interactions, and Variability. Antarctic Research Series 73, American Geophysical Union, Washington DC.

Longhurst, A.R. (1998) Ecological Geography of the Sea. Academic Press, Florida.

Lyne, V and Hayes, D. (2005) Pelagic Regionalisation: National Marine Bioregionalisation Integration Project. CSIRO Report to the National Oceans Offi ce. CSIRO Marine Research, Hobart. (Summary)

Massom, R. A. (2003) Recent iceberg calving events in the Ninnis Glacier region, East Antarctica, Ant. Sc., 15(2): 303-313

Massom, R. A., Harris, P.T., Michael, K.J. and Potter, M.J. (1998) The distribution and formative processes of latent-heat polynyas in East Antarctica. Ann. Glaciol., 27: 420-426 Designed and produced by Fresco Creative: www.frescocreative.com.au Creative: by Fresco Designed and produced

Bioregionalisation of the Southern Ocean

Report of Experts Workshop

(Hobart, September 2006)

APPENDICES

Appendix I: Approaches to bioregionalisation – examples presented during the workshop

Antarctic Environmental Domains Analysis

CCAMLR Small-Scale Management Units for the fishery on Antarctic krill in the SW Atlantic

Australian National Bioregionalisation: Pelagic Regionalisation

Selecting Marine Protected Areas in New Zealand’s EEZ

Appendix II: Technical information on approach to bioregionalisation

Appendix III: Descriptions of datasets used in the analysis

Appendix IV: Results of secondary regionalisation using ice and chlorophyll data

Appendix V: Biological datasets of potential use in further bioregionalisation work

Appendix VI: Details of datasets, Matlab code and ArcGIS shapefiles included on the CD Appendix I

Approaches to bioregionalisation: examples presented during the workshop

Example 1: Antarctic Environmental Domains Analysis Harry Keys and Fraser Morgan (Department of Conservation, New Zealand)

Annex V of the Protocol on Environmental environment of the continent included climate Protection to the Antarctic Treaty states that (mean annual temperature, seasonal “Parties shall seek to identify, within a temperature range, mean annual wind speed systematic environmental-geographic and average atmospheric solar radiation), framework…” a series of Antarctic Specially diurnal length, slope, ice cover, land-form and Protected Areas. New Zealand has been geology. working with the CEP on the elaboration of a systematic environmental geographic The “proof of concept” classification provides framework (SEGF), designed to provide a a scientifically sound basis for differentiating spatial classification for a range of parts of the ice sheet, and ground truthing management activities including identification shows that the classification is realistic for the of priority sites for protection, environmental ice sheet. The 20-group classification was protection, and assessment of risks also tested using data included in protected associated with human activities. The work area management plans (including biotic programme and review have been greatly data), and was found to be consistent with assisted by Australia, the Russian Federation, these descriptions. However further testing is the British Antarctic Survey and the Scientific still needed using biological and geological Committee for Antarctic Research. datasets. The most recent version of the classification (version 1.1) is shown in Figure Environmental Domains Analysis has been A1. A finer scale classification layer used to create a “proof of concept” comprising perhaps 100 Environments will be classification of the Antarctic continent into 20 required at regional scales or areas of high Environments (Morgan et al., 2005). Data spatial variability such as ice-free land layers used to differentiate the physical environments.

FIGURE A1: Environmental Domains of Antarctica (version 1.1, Landcare Research)

i Example 2: CCAMLR Small-Scale Management Units for the fishery on Antarctic krill in the Southwest Atlantic Roger Hewitt (NOAA, USA)

Following an international, multi-ship survey in where the fishery obtains relatively reliable the Southwest Atlantic region in January– catches over a number of years). Such February 2000, the biomass of Antarctic krill locations were identified by their relative (Euphausia superba) in this region was within year importance averaged over several estimated to be 44.3 million tonnes and a new years. Computational steps were: 1) bin the precautionary catch limit of 4.0 million tonnes data over a fine-scale geographic grid; 2) was adopted by CCAMLR for FAO Statistical normalize the data within each year; 3) Area 48 (Southwest Atlantic) (Hewitt et al., geographically smooth the data within each 2002). The Commission further subdivided the year; 4) average values over time series to catch limit among four Subareas, in order to obtain average importance of locations; and 5) distribute fishing effort and thereby reduce the identify area of high importance by applying a potential impact of fishing on land-based threshold. In synthesizing the results, the predators. However, concern remained that workshop first separated pelagic areas from localised depletion of the krill resource could those areas considered important to colonies occur if all the allowed catch was taken within of land-based predators (as defined by a small proportion of a Subarea. The maximum foraging distance). A second Commission therefore required that the total division was made based on common patterns catch in all of Area 48 should not exceed of krill aggregation, predator foraging and 620,000 tonnes until the precautionary catch fishing. limits for each Subarea had been subdivided into small-scale management units (SSMUs). The recommendations of workshop were A workshop was convened to define SSMUs adopted by the Commission in 2002, which in Area 48 (SC-CAMLR 2002), based on then asked for advice on how the considerations of land-based predator precautionary catch limit for krill in Area 48 foraging ranges, krill distribution and the could be subdivided among the SSMUs. behaviour of krill fishing vessels (Constable Several options have been outlined (Hewitt et and Nicol, 2002). al. 2004), and work is ongoing to provide specific advice to the Commission. The workshop defined common patterns among krill aggregations (defined as Summary plots for the South Shetland Islands predictable locations where krill are found at (Subarea 48.1) are shown in Figure A2. The relatively high densities over a number of scheme of 15 SSMUs (12 coastal and 3 years), predator foraging areas (defined as pelagic) adopted by the Commission for predictable locations where a predator obtains Subareas 48.1, 48.2 and 48.3 are shown in food over a number of years), and fishing Figure A3. grounds (defined as predictable locations

ii

60.0 60.0 January Surveys (1998, 1999, 2001) January Surveys (1998, 1999, 2001) Krill NASC (normalized) Krill NASC (normalized) Step 0.1, distance 1, smoothing 0.1 Step 0.1, distance 1, smoothing 0.1 60.5 60.5 Krill fishery 61.0 Krill 61.0

61.5 61.5 e e d d tu tu ti ti a a L L 62.0 62.0

62.5 62.5

63.0 63.0 o 60 S 63.5 63.5 63.0 62.0 61.0 60.0 59.0 58.0 57.0 56.0 55.0 54.0 63.0 62.0 61.0 60.0 59.0 58.0 57.0 56.0 55.0 54.0 Longitude Longitude

60.5 -60.5 South Shetlands Finfish Distribution (krill predators) 1998 and 2001

61 -61.0 Land-based krill o Krill eating 62 S predators 61.5 -61.5 fish e d

ude t tu

i 62 t

-62ti .0 a La L

-62.5 62.5 o 64 S -63.0 63

-63.5 63.5 63 62 61 60 59 58 57 56 55 54 53 -62 -60 -58 -56 -54 Longitude Longitude o 66 S FIGURE A2: Delineation of Small-Scale Management Units in the vicinity of the South Shetland Islands (CCAMLR Subarea 48.1) based on time-averaged patterns of krill o distribution, predator foraging and krill fishing. 68 S

o o o o o 66 W 63 W 60 W 57 W 54o 69 W W

FIGURE A3: Small-Scale Management Units (SSMUs) for the fishery on Antarctic krill in CCAMLR Subareas 48.1, 48.2 and 48.3. The scheme is composed of 15 SSMUs, three pelagic (Antarctic Peninsula (APPA), South Orkney Islands (SOPA), and South Georgia (SGPA), and 12 coastal.

iii Example 3: Australian National Bioregionalisation: Pelagic Regionalisation Vincent Lyne and Donna Hayes (Department of the Environment and Heritage and CSIRO MAR) (Summary from Lyne and Hayes, 2005)

The pelagic regionalisation component of the 4. A workable pelagic framework has been Australian National Marine Bioregionalisation constructed down to the scale of features was developed using a hierarchical framework (Level 4a) and substructure of features (Level that relies primarily on physical properties and 4b). satellite plankton images. Three levels of the framework (Levels 1–3) are described by 5. Beyond Level 2, energetics in the ocean application to Australia’s Marine Jurisdiction. system as characterised by fields of The first level (Level 0) describes the structure homogeneity and heterogeneity in water at the scale of the oceans. The second level masses provide useful qualifiers for Level 2 (Level 1) describes ocean zones that appear classes. The rich variety of oceanographic as latitudinal bands at the surface (Level 1a) processes that can be identified with such an but also have an underlying three-dimensional analysis may have implications for biological structure (Level 1b). The third level (Level 2) productivity and hence for marine resource describes the different circulation regimes management. Deciding how best to use the arising from mixing and transport of water available information at the various Levels properties. The fourth level (Level 3) and what future information will help refine the describes the energetics and variability of analyses and descriptions will require careful water masses; we present only the surface consideration. characterisation at this level. Although finer- At this stage, the pelagic regionalisation of scale levels (Levels 4a and 4b) are presented Australia is being used mainly to illustrate the in the regionalisation framework, these are complexity of the structure of the marine water dynamic features whose description is left as column. It is at an earlier stage of a subject for future research. development than the benthic regionalisation. Key conclusions of this project are However, as with the benthic regionalisation, summarised as follows: the key aim ultimately in using the pelagic regionalisation is to determine linkages 1. Substantial progress has been made in between biological and geophysical attributes, classifying the pelagic environment through an which we expect to be much stronger in integrated analysis of the whole water column. pelagic than benthic systems. Our work also A hierarchically nested classification appears suggests that the coupling between pelagic possible from the scale of oceans (Level 0) and topographic structures is much tighter down to at least Level 2 (Circulation than we expected, and that a unified treatment Regimes). However, the complications due to of pelagic and benthic regionalisations should seasonal and longer timescale variability are be possible. A corollary conclusion is that yet to be examined in a systematic way. benthic structures may be influenced by the pelagic environment. 2. Biological information is required to guide and inform the analyses; this is currently the Figure A4 shows the three-dimensional nature main major shortcoming of the regionalisation. of water masses defined in the pelagic regionalisation. 3. This is one of the most comprehensive studies undertaken for a pelagic regionalisation. A valuable collection of data for future extensions of this study now exists in a well documented state (Hayes et al., 2005).

iv

FIGURE A4: Image demonstrating the three-dimensional nature of water masses defined in the Australian National Bioregionalisation pelagic regionalisation. For example, water mass 13 in the Pacific Ocean is the Pacific Central-South subtropical water (coloured green) and extends from the surface waters to around 250 m depth. Water mass 10 (the southern Subtropical Convergence – coloured light blue) is part of the Southern Ocean, and extends from the surface to around 800 m depth. (Image produced using a demonstration version of Amira). (Reproduced with permission from: Commonwealth of Australia (2005) National Marine Bioregionalisation of Australia. Summary. Department of Environment and Heritage, Canberra, Australia).

v Example 4: Selecting marine protected areas in New Zealand’s EEZ John Leathwick (NIWA, New Zealand)

Prior to 2003, we developed an environmental classification describe the distributions of 122 classification of New Zealand’s exclusive demersal fish species for cells in 1.9 million 1 economic zone (EEZ) for use as a generic km2 grid cells across New Zealand’s EEZ. classification for management of marine These layers were predicted from statistical conservation issues. During the first half of models fitted using boosted regression trees 2006 we developed two alternative and relating the abundance of individual fish management classifications for the EEZ and species to environment, using data from then explored the use of reserve selection 17,000 research trawls. The predictive software, this research being funded jointly by accuracy of these models was assessed by New Zealand’s Department of Conservation predicting to an independent set of 4100 and Ministry of Fisheries. trawls, and comparing predicted and actual catch. The classification was defined using a In the first phase of this work we produced an similar clustering procedure as used above to environmental classification that was define the fish-tuned environmental specifically tuned for use in managing issues classification, and gave a further significant related to demersal fish (Leathwick et al., gain in classification strength compared to 2006a). For this classification, we used data both the original generic environmental from 17,000 research trawls to guide both our classification and our fish-tuned selection of variables to include, and their environmental classification. weighting and transformation. The classification was defined using an initial non- In the third phase of this work, we used the hierarchical classification to define around 100 reserve selection software Zonation (Molianen groups, and the centroids from these groups in press) to explore possible marine protected were then used to define a hierarchical area configurations for New Zealand’s EEZ classification. This classification had (Leathwick et al., 2006c). This analysis used significantly better classification strength than the same predicted demersal fish distribution our original generic environmental layers as used to produce the fish classification, i.e. it had greater ability to act assemblage classification in phase two. It as a surrogate for variation in fish community produces a conservation ranking for all grid composition as described by fish catch data cells – designation of the highest-ranked cells from research trawls. as marine protected areas would achieve the highest levels of conservation protection, In the second classification, rather than averaged across all species. Results indicate defining an environmental classification, we that setting aside 10% of New Zealand’s EEZ used a set of biological layers to define a would protect, on average, around 30% of biological classification that describes directly species’ geographic ranges. Increasing the the geographic distribution of fish geographic extent of marine protected areas assemblages (Leathwick et al., 2006b) (Figure to 20% would increase the average level of A5). The input data layers for this species range protection to nearly 50%.

vi

FIGURE A5: Geographic distribution of 16 fish assemblages, defined on the basis of similarity in fish composition for sites with an average depth less than the maximum depth recorded in the fish_comm. research trawl database (1950 m). (Reproduced with permission from: Leathwick et al., 2006b).

vii Appendix II

Technical information on approach to bioregionalisation

Choice of algorithm and summary parameters

The choice of statistical clustering method processing steps to take, and so on) creates (algorithm) and the parameters used to uncertainty in the final output. Different characterise ecosystem features can influence choices at the different processing stages can the results of the bioregionalisation analysis. lead to quite different region boundaries. This The choice of algorithm usually involves a type of uncertainty is distinct from the trade-off between algorithm assumptions, variability that arises from spatial, temporal, complexity (computational time), accuracy, and other stochastic variations in ecosystem and other factors. There may also be several processes. Epistemic uncertainty can appropriate algorithms for a given application, potentially be reduced through an improved each with slightly different trade-offs and understanding of the relevant ecosystem giving differing results. Clustering algorithms processes. However, there will always be can be broadly categorised as dissimilarity- some level of uncertainty in the final result. based or model-based; both categories can Most end-use applications of a regionalisation pose difficulties in combining variables of require an understanding of the limitations different types. A dissimilarity metric or model inherent in that regionalisation. structure that is suitable for one type of data (e.g. sea surface temperature, measured on a Scaling and weighting variables continuous scale) may not be appropriate for another type (e.g. ocean current, measured as speed and direction). Variables of comparable The data inputs to the regionalisation process type but measured on different scales are represent a collection of variables that differ in often normalised (e.g. by standardising the their measurement units and in the way those mean and variance, or range of each variable) units represent their interactions with the – again there are a number of normalisation ecosystem. In order to be able to combine algorithms, often with limited theoretical these variables in a statistical model, each guidance but considerable potential variation variable must be normalised so that the units in results between the different algorithms. of one variable are directly comparable with Variables that are influenced by multiple those of another. A variety of normalisation ecological processes will reflect different schemes are possible. The simplest schemes aspects of those processes depending on how simply carry out a linear rescaling of each that variable is incorporated into the analyses. variable’s values, so that they are in some sense standardised (e.g. so that each variable’s values lie between zero and one, or Spatial and temporal scales so that each variable has a mean of zero and standard deviation of one). Linear methods The spatial and temporal scale of the data such as these run into difficulties in situations should be appropriate to the desired scale of where the ecological relevance of the variable the regions. Data with fine-scale structure may is nonlinear. A classical example of this is need to be smoothed for use in broad-scale water depth – a change of depth from 50 to regionalisations. For pelagic applications, the 100 m is likely to be associated with a much selection of spatial regions is complicated by larger change in ecosystem properties than a the depth structure of the water column and change in depth from 2000 to 2050 m. temporal variability at seasonal and longer Nonlinear scaling methods can incorporate timescales. A hierarchical approach is often this type of nonlinear relevance by used to assist in resolving problems of scale. emphasising those parts of a variable’s range The levels of the hierarchy can represent that have the maximum ecological relevance. either spatial scales, or different ecological Nonlinear normalisation methods include processes. A process-oriented hierarchy often histogram scaling (replacing the variable’s has an approximate spatial structure due to values by their ranks; Lyne et al. 2005) and the spatial scales of the processes. other commonly used transformations such as log scaling. Thresholding can also be used, Uncertainty wherein a variable is broken into a series of ranges, and within each range the variable has approximately constant ecological effect. Incomplete knowledge (epistemic uncertainty) The range thresholds are chosen to match about how the regionalisation process should known biological thresholds (e.g. the shelf be carried out (which data to choose, what break is generally considered to be between

viii 500 m and 1000 m around Antarctica, so a spatially separated. A region is thus threshold of 500 m might be suitable for a considered to be a group of sites that belong depth variable). to the same cluster, but which also form a contiguous spatial area. A single cluster may The choice of normalisation scheme is rarely produce a number of regions, each of which clear-cut, but should obviously be guided by have the same general properties, but which expert knowledge as to the role of each are spatially distinct. variable in the ecosystem at the scales of interest to the regionalisation exercise. Clustering algorithms are often based around Statistical methods such as regression trees the concept of a dissimilarity metric, which (in can be used to guide or even perform the the context of a regionalisation) is used to normalisation process. These approaches calculate how dissimilar two sites are, given generally seek to find a normalisation scheme their ecosystem properties (physical or that matches the changes in the variable to biological data). The clustering of sites into those in a dependent (usually biotic) variable. regions is carried out in such a way that the intra-region dissimilarity of sites is low (i.e. The contribution of each variable to the final sites within a region are similar to each other) regionalisation also needs to be considered. and inter-region dissimilarities are high. An Most clustering algorithms will produce an alternative class of clustering algorithms – outcome that reflects the information present known as model-based clustering – do not in each of the input variables with an equal use a dissimilarity metric in this way, but weighting. In some cases it might be instead attempt to model the data density and appropriate to upweight the contribution of then use the properties of the fitted model to one or more variables, if they are seen to be decide how the sites should be partitioned into ecologically more relevant than the others in regions. Model-based approaches can the input data. Conversely, two or more become cumbersome when applied to variables that are correlated (in the sense that mixtures of different types of variables (e.g. a they each reflect aspects of the same continuous variable such as sea surface underlying ecological process) might need to temperature with a discrete variable such as be downweighted in order to avoid an undue the presence or absence of a particular emphasis on that process. Statistical methods species). For simplicity and flexibility we focus can again be used to guide weighting here on dissimilarity-based methods and do schemes. Variable selection methods might not consider model-based approaches further. also be useful in deciding which input variables should be chosen for a given Dissimilarity-based clustering methods can be regionalisation. broadly divided into hierarchical or non- hierarchical schemes. Hierarchical Classification methods agglomerative algorithms take an iterative approach in which all sites are initially allocated to their own region. At each The intent of a regionalisation is to partition iteration, the two most similar regions are the study area into a set of discrete spatial merged together, until at the end of the regions, each with relatively homogeneous process there is only one region (which ecosystem properties. Clustering algorithms contains all of the sites). This hierarchy are well suited to this task, as they are (termed a dendrogram) can then be cut at an designed to partition a large data set into a appropriate level to give the desired number number of subsets, each with relatively similar of regions. Hierarchical methods are generally properties that differ from those of the other computationally expensive, but provide a subsets, In the context of a regionalisation, considerable advantage over non-hierarchical the clustering process takes sites from a methods in that the dendrogram can be used regular grid in geographic space. Each site to examine the relationships between the has associated ecological properties (physical different regions that are produced. The and/or biotic data) and this information is used unweighted pair-group method with arithmetic to group together sites with relatively similar mean (UPGMA) is a very commonly used ecological properties. Those groupings (which hierarchical method. are calculated only on ecological properties, not spatial information) are then projected Non-hierarchical methods generally operate back into geographic space in order to find the by firstly selecting a number of sites as initial spatial extents of the resulting regions. cluster centres. Each site is then allocated to the cluster to which it is most environmentally It is important to make the distinction between similar. An iterative process is then used to the clusters that are produced by a clustering repeatedly calculate the environmental algorithm, and the regions that are formed “centroid” of each cluster, and re-allocate sites from those clusters. A cluster is a group of to the most environmentally similar cluster, sites that are considered to have similar until convergence is reached. Non-hierarchical ecological properties. However, because the algorithms are generally less computationally clustering process is based on environmental demanding than hierarchical ones and so can similarity (and not spatial information), a be used with larger data sets. single cluster may contain sites that are

ix Dissimilarity-based clustering algorithms advantage of the computational efficiency of require an appropriate dissimilarity metric to the non-hierarchical approach to reduce the be chosen. Briefly, the dissimilarity metric full set of grid cells down to a manageable estimates the environmental dissimilarity of number, which can then be analysed two sites, given a set of environmental data hierarchically. This yields a final dendrogram for each. Dissimilarity metrics have been the and a major advantage in the interpretability subject of ongoing research in numerical of the final cluster set. This hybrid approach is ecology; the reader is referred to (for identical to that used by Leathwick et al., example) Faith et al. (1987). These analyses 2003. used the Gower metric, which for sites j and k with data vectors x and x is calculated: The choice of the number of clusters for the j k non-hierarchical output is somewhat arbitrary djk = ∑i |xij – xik|/(max(Xi) - min(Xi)) – it should be much greater than the final number of clusters expected from the where max(Xi) and min(Xi) are the maximum hierarchical step, but not so many that the and minimum values of variable i across the hierarchical algorithm becomes too slow. We whole study area. The Gower dissimilarity is used 40 clusters, giving a fast computation effectively the proportional difference in time, but for a production run this should environment between the two sites, across all probably be increased to 200 or more (thus variables. retaining more of the detail present in the full A further aspect of the clustering process that set of sites). The non-hierarchical algorithm must be explored is the choice of the number used was the CLARA clustering routine of clusters. Most clustering algorithms can (Kaufman & Rousseeuw 1990) with a Gower produce somewhere between one cluster (i.e. distance metric. The Gower distance was in all sites belong to the same region) and as fact implemented by first range-standardising many clusters as there are sites (each each variable (so that each variable’s values belongs to its own region). The optimal choice were between zero and one) and then between these two extremes will be governed applying a Manhattan (city-block) distance by the intended ecological and spatial scale of metric. The CLARA routine uses a the regionalisation. A broad-scale subsampling strategy to deal with large data regionalisation – one that is intended to reflect sets. fairly high-level ecological processes – would The hierarchical clustering phase used the require a smaller number of clusters than a common unweighted pair-groups method with regionalisation that is intended to focus on arithmetic mean (UPGMA; see e.g. Kaufman small-scale processes. & Rousseeuw 1990), again with a Gower metric. Adopted method The analyses were carried out using custom- th written scripts in Matlab (v7.1; Mathworks, The classifications were performed on a 1/8 Natick, MA, 2006). The CLARA clustering degree grid, with a total of 901,440 grid cells used was that implemented in the cluster (including land) from 80°S to 40°S. A land package of the R statistical package mask was derived from the GEBCO (http://cran.r-project.org/), called from within bathymetry data; any grid cell with a mean Matlab using the R-DCOM server. The depth greater or equal to zero was marked as UPGMA clustering was that provided as part land. Land pixels were removed (leaving of Richard Strauss’ Matlab toolbox 720,835 cells) before performing the (http://www.biol.ttu.edu/Strauss/Matlab/matlab classifications. .htm). Maps were produced within Matlab The classification method adopted during the using the m_map toolbox (Rich Pawlowicz; workshop was a mixed non-hierarchical and http://www.eos.ubc.ca/~rich/map.html) with hierarchical method. The full set of grid cells final maps produced in ArcGIS (version 9). was subjected to a non-hierarchical clustering The Matlab scripts used in the workshop to produce 40 clusters. The mean data values analysis are included on this CD (see for each of the 40 clusters was calculated and Appendix VI for details). then a hierarchical classification was performed to produce a dendrogram and the final clustering. This approach takes

x Appendix III

Description of datasets used in the analysis

Sea surface temperature (SST)

Mean annual sea surface temperatures were obtained from the NOAA Pathfinder satellite annual climatology (Casey and Cornillon 1999). This climatology was calculated over the period 1985–1997 on a global 9km grid. Monthly values were averaged to obtain an annual climatology. Casey, K.S. and P. Cornillon (1999) A comparison of satellite and in situ based sea surface temperature climatologies, J. Climate, vol. 12, no. 6, pp. 1848-1863. http://podaac-www.jpl.nasa.gov/products/product112.html

Bathymetry

Depth data were obtained from the GEBCO digital atlas (IOC, IHO and BODC, 2003). These data give water depth in metres and are provided on a 1-minute global grid. Centenary Edition of the GEBCO Digital Atlas, published on CD-ROM on behalf of the Intergovernmental Oceanographic Commission and the International Hydrographic Organization as part of the General Bathymetric Chart of the Oceans, British Oceanographic Data Centre, Liverpool, U.K. See http://www.gebco.net and http://www.bodc.ac.uk/projects/international/gebco/ A metadata record can be obtained from: http://aadc- maps.aad.gov.au/aadc/metadata/metadata_redirect.cfm?md=AMD/AU/gebco_bathy_polygons

Nutrient concentrations

Silicate and nitrate concentrations were obtained from the WOCE global hydrographic climatology (Gouretski and Koltermann, 2004). This climatology provides oceanographic data on a 0.5º regular grid on a set of 45 standard levels covering the depth range from the sea surface to 6000m. The silicate and nitrate concentrations were calculated from seawater samples collected using bottles from stationary ships. The nutrient concentrations at the 200m depth level were used here; concentrations are expressed in µmol/kg. Gouretski, V.V., and K.P. Koltermann, 2004: WOCE Global Hydrographic Climatology. Technical Report, 35, Berichte des Bundesamtes für Seeschifffahrt und Hydrographie. http://odv.awi-bremerhaven.de/data/ocean/woce-global-hydrographic-climatology.html

Insolation (PAR)

The mean summer climatology of the photosynthetically-active radiation (PAR) at the ocean surface was obtained from satellite estimates (Frouin et al.). These PAR estimates are obtained from visible wavelengths and so are not available over cloud- or ice-covered water, or in low-light conditions including the austral winter. Hence in the sea ice zone, this climatology represents the average PAR calculated over the period for which the water was not ice-covered. Robert Frouin, Bryan Franz, and Menghua Wang. Algorithm to estimate PAR from SeaWiFS data Version 1.2 - Documentation. http://oceancolor.gsfc.nasa.gov/cgi/climatologies.pl?TYP=swpar

Chlorophyll-a

Mean summer surface chlorophyll-a concentrations were calculated from the SeaWiFS summer means. We used the mean of the 1998–2004 summer values. Chlorophyll concentrations are expressed in mg/m3. http://oceancolor.gsfc.nasa.gov/cgi/level3.pl

xi Sea ice

We calculated the mean fraction (0-1) of the year for which the ocean was covered by at least 15% sea ice. These calculations were based on satellite-derived estimates of sea ice concentration spanning 1979-2003. Comiso, J. (1999, updated 2005). Bootstrap sea ice concentrations for NIMBUS-7 SMMR and DMSP SSM/I. Boulder, CO, USA: National Snow and Ice Data Center. Digital media. http://nsidc.org/data/nsidc-0079.html

Southern Ocean Fronts

These are the front positions as published by Orsi et al. (1995). Orsi A, Whitworth T, III, Nowlin WD, Jr (1995) On the meridional extent and fronts of the Antarctic Circumpolar Current. Deep-Sea Research 42:641-673 A metadata record can be found at: http://aadc- maps.aad.gov.au/aadc/metadata/metadata_redirect.cfm?md=AMD/AU/southern_ocean_fronts

xii Appendix IV

Results of secondary regionalisations using ice and chlorophyll data

This section presents the results of the exploratory secondary regionalisations undertaken using the primary variables plus additional data on ice and chlorophyll. These secondary variables have particular relevance in the coastal and seasonal sea ice zones, and may be of value in future, fine- scale classifications for these areas.

Ice

The Antarctic coastal marine zone is complex, Sea ice forms an important habitat for algae, and strongly influenced by the dynamics and bacteria and invertebrates such as krill, thermodynamics of sea ice (see Section 3.1 in copepods and salps living under and within the main report). Sea ice distribution and the ice, and plays a dominant defining role in characteristics (of both pack and fast ice) are structuring high-latitude marine ecosystems strongly affected by coastal configuration and on a variety of scales, particularly in the the presence of grounded icebergs, which are Seasonal Ice Zone (the area between the in turn closely linked to bathymetry (Massom maximum northern extent of sea ice in winter et al., 2001). and minimum in summer). On the continental shelf area, sea ice forms a foraging, breeding The processes that may drive regionalisation and resting habitat for pack-ice seals, in this area include the annual advance and penguins and land-based predators (Ainley et retreat of sea ice, the persistence of perennial al., 2003). Fast ice, which forms in sheltered (multi-year) ice, the formation of polynyas and bays and adjacent to grounded icebergs, also fast ice, and the build-up of highly-deformed constitutes a key habitat for breeding Weddell ice adjacent to iceberg grounding zones. seals (Leptonychotes weddellii) and Emperor Coastal currents (including the large-scale penguins (Aptenodytes forsteri). Adjacent Antarctic Coastal Current), bathymetry and polynyas and marginal ice zones (within which offshore (katabatic) and prevailing easterly wave-ice interaction is a key process) also winds influence the concentration and extent form a critical foraging habitat for many of sea ice and the movement of icebergs. species of birds and marine mammals. In Coastal polynyas (large recurrent areas of addition, the ice edge is typically a region of open water/thin ice) occur all around the enhanced biological activity during the melt continent, while deep water (offshore) season in particular (Nicol and Allison, 1997; polynyas occur in two locations (the Weddell Smith and Nelson, 1986; Smith et al., 1988; and Cosmonaut seas). Coastal “latent heat” Sullivan et al., 1993). polynyas constitute major regional sea ice “factories”, sites of major water-mass Some of the processes and features modification (Bindoff et al., 2000) and, in associated with sea ice dynamics in the places, regions of enhanced biological coastal zone are illustrated in Figure A6. activity, and usually have a similar location and extent from year to year. By comparison, A secondary regionalisation was undertaken the formation of the deep water polynyas, using data on the primary variables (depth, which are regions of intense ocean- sea surface temperature, nitrate, silicate) and atmosphere interaction, tends to be more the proportion of the year for which an area is sporadic (Morales Maqueda et al., 2004). covered by at least 15% concentration of sea ice. The results of this classification are Icebergs may have a significant impact on the shown in Figure A7. The defined regions marine environment through scouring and identify similar general features to those grounding. These processes have particular shown in the primary regionalisation (Figure significance in the continental shelf area, 18), however more heterogeneity can be seen where water depth is less than 500 m in the coastal and shelf areas, which may (Beaman and Harris, 2005). While large correspond to the types of features illustrated icebergs attract attention, the coastal zone in Figure A6. also contains many thousands of small grounded icebergs (Massom, 2003; Young et al., 2006).

xiii

DYNAMIC ELEMENTS Seasonal sea ice zone in coastal zone Coastal current Offshore wind Islands Fast ice Continental shelf edge Polynya

EWD Residual pack ice LBMP breeding

Bank (<400m) with icebergs Continental Glacier tongue coastline Island Ice shelf Ice shelf

FIGURE A6: Preliminary conceptual diagram of the Antarctic coastal marine zone and processes that may drive regionalisation. Basic elements are the coastline, the continental shelf edge, and islands. Other fixed features are also labelled. The LBMP breeding symbols refer to the locations of land- based marine predator colonies, which includes all birds and mammals that breed on land or ice. Although often in a fixed location, these colonies are seasonally dynamic, and these symbols are indicative only. Residual pack ice in this case refers to regions of perennial sea ice i.e., sea ice that survives the summer melt season. Figure developed by H. Keys and R. Massom, based on other work including Beaman and Harris, 2005.

xiv 30°W 0° 30°E

40°S 40°S

90°W 90°E

40°S 40°S

0 500 1000 2000 Km

Projection: Polar stereographic True scale at 71°S

150°W 180° 150°E Figure A7: Secondary regionalisation of the Southern Ocean using depth, sea surface temperature, silicate and nitrate concentrations (primary variables) plus the proportion of the year for which areas are covered by at least 15% concentration of sea ice (secondary variable). (12 clusters)

(Note that the colours of the regions in this classification do not correspond to those used in other classifications).

Regions 1 4 7 10 2 5 8 11 3 6 9 12 Chlorophyll

A further secondary regionalisation was abundance, the phytoplankton accumulate to undertaken using the primary properties plus high levels that are detected by satellite data on the chlorophyll a concentration. images as chlorophyll a. In a different location Chlorophyll a concentration is highly where herbivory is high, the phytoplankton are seasonal, and also has a high degree of inter- highly productive but are eaten by herbivores annual variability. Satellite-derived chlorophyll as fast as they appear, hence chlorophyll a a concentration can be used as a proxy for remains low. These issues may need to be phytoplankton distribution, although it may not considered further in future analyses where be an accurate reflection of primary information on the abundance of production. phytoplankton and herbivores is of particular importance. However, for the purposes of this Phytoplankton abundance is a function of two analysis, chlorophyll a provides a useful proxy dynamic processes: production and herbivory. for exploring spatial heterogeneity in primary Chlorophyll a concentration can provide a production. good estimate of the distribution and biomass of phytoplankton at any particular moment, Figure A8 shows the secondary and may also provide an indication of the regionalisation using primary variables plus distribution of herbivores. However, two areas chlorophyll a. A higher level of heterogeneity with identical phytoplankton productivity may than in the primary regionalisation is have very different phytoplankton particularly evident in the coastal and shelf abundances. In a location with low herbivore areas, and also around island groups.

Ice and chlorophyll

A final secondary classification was balance between the influence of these two undertaken using both ice and chlorophyll secondary variables. Some of the complexity data (in addition to the primary variables). A evident particularly in Figures A8 and A9 may large number of clusters (40) were used to also be due to subsurface current dynamics, illustrate the high level of heterogeneity which may affect silicate and nitrate generated by variation in these properties. concentrations. This secondary regionalisation is shown in The complexity of the secondary Figures 21, 23 and 25 for each of the regionalisations shown here highlights the CCAMLR statistical areas, and is also shown need for further work to identify the in Figure A9 for the Southern Ocean as a appropriate level of regional separation at whole. Chlorophyll a is likely to be driving smaller scales, using these secondary much of the heterogeneity across the open datasets. ocean areas of the Southern Ocean in this regionalisation, whereas in the coastal and seasonal ice zones there may be more of a

xvi 30°W 0° 30°E

40°S 40°S

90°W 90°E

40°S 40°S

0 500 1000 2000 Km

Projection: Polar stereographic True scale at 71°S

150°W 180° 150°E Figure A8: Secondary regionalisation of the Southern Ocean using depth, sea surface temperature, silicate and nitrate concentrations (primary variables) plus chlorophyll a concentration (secondary variable). (15 clusters)

(Note that the colours of the regions in this classification do not correspond to those used in other classifications). Regions 1 5 9 13 2 6 10 14 3 7 11 15 4 8 12 30°W 0° 30°E

40°S 40°S

90°W 90°E

40°S 40°S

0 500 1000 2000 Km

Projection: Polar stereographic True scale at 71°S

150°W 180° 150°E Figure A9: Secondary regionalisation of the Southern Ocean using depth, sea surface temperature, silicate and nitrate concentrations(primary variables) plus the proportion of the year for which areas are covered by at least 15% concentration of sea ice, and chlorophyll a concentration (secondary variables). (40 clusters)

(Note that the colours of the regions in this classification do not correspond to those used in other classifications). Regions 1 5 9 13 17 21 25 29 33 37 2 6 10 14 18 22 26 30 34 38 3 7 11 15 19 23 27 31 35 39 4 8 12 16 20 24 28 32 36 40 Appendix V

Biological datasets of potential use in further bioregionalisation work

Several biological datasets were identified by participants during the course of the workshop as having potential relevance in future bioregionalisation work. These are listed in Table A1. This is not an exhaustive list, and further work is required to identify and collate biological datasets, and to prioritise which of these datasets may be of use in future, fine-scale bioregionalisation analyses. A large amount of data is likely to be available through national Antarctic programmes, which may have specific relevance to particular areas of the Southern Ocean. Such datasets have not been explored in detail here. Other international organisations may also have relevant datasets for areas overlapping with, or adjacent to the Southern Ocean, for example other Regional Fisheries Management Organisations, and the Agreement on the Conservation of Albatrosses and Petrels (ACAP).

TABLE A1: Biological datasets of potential use in further bioregionalisation analysis Data source Reference Notes SCAR MarBIN http://www.scarmarbin.be/ Integrated database system (Marine Biodiversity including distribution records Information on invertebrates, fish, birds, Network) seals and whales from a range of sources. Southern Ocean http://aadc- Quantitative data, analysed in a Continuous maps.aad.gov.au/aadc/cpr/index.cfm consistent matter, across Plankton Recorder latitudinal transects on voyages (SO-CPR) Survey (also available through SCAR MarBIN) to/from Antarctica. 14,000 and SCAR Action records. Group on CPR Research Includes zooplankton to species level, plus simultaneous water column characteristics, including fluorometry, salinity, temperature, and incident light Spatial distribution is biased toward the Australian sector, with limited CPR tows around the rest of the Southern Ocean. Southern Ocean http://www.antarctica.ac.uk/BAS_Science Comprehensive distribution Mollusc Database /programmes 2000-2005/ABPPF/ records of Antarctic, (SOMBASE) SOMBASE/index.php Magellanic, and Sub-Antarctic Gastropods and Bivalves. (also available through SCAR MarBIN) Distribution maps form part of a biogeographic database, which also includes taxonomic, ecological and habitat data. Database contains information on over 1,400 species from more than 3,350 locations. IWC historical (available through SCAR MarBIN) Spatially comprehensive; catch whaling / catch data records for austral summers from 1931/32 to 1979/80

xix Data source Reference Notes CCAMLR datasets (access according to CCAMLR data access Fish catch data rules) Observer data (includes bycatch plus bird and mammal sightings) Acoustic (krill) survey data CCAMLR 2000 Survey (synoptic data on oceanography, zooplankton, krill and higher predator biomass for the Scotia Sea and Antarctic Peninsula region) Antarctic Pack Ice http://nmml.afsc.noaa.gov/apis/apis.htm Multi-national survey data on Seals data (SCAR crabeater, leopard, Weddell, Group of Specialists (access through SCAR Group of Specialists and Ross seals on Seals) on Seals)

BirdLife International http://www.birdlife.org/action/science/ Tracking Ocean Wanderers – procellariiform species/seabirds/tracking_ global distribution data on tracking data ocean_wanderers.pdf albatrosses and petrels (access according to BirdLife data access rules) Seabird colony (access through SCAR Birds group) Geographic information on locations (Antarctic colony locations and size (data Peninsula region) see also: used in Wildlife Awareness http://www.era.gs/resources/wam/WAMv3- Manual complied from existing FINAL-web.pdf sources – see references in publication) Seabird colony (access through SCAR birds group) Some datasets available from locations (East Australian Antarctic Data Antarctica, and see also: Centre Portal other continental, http://aadc-maps.aad.gov.au/aadc/portal/ Antarctic Peninsula and high latitude subantarctic islands records ) ‘Sea Around Us’ http://www.seaaroundus.org/lme/ Includes data on marine fish portal (UBC) – SummaryInfo.aspx?LME=61# (FishBase) and global Antarctic Large database of potential Marine Ecosystem http://www.seaaroundus.org/global/ seamounts Seamounts Online http://seamounts.sdsc.edu/ Taxonomic and distribution data on seamount species worldwide

xx Appendix VI

Description of datasets, Matlab code, and ArcGIS shapefiles included on the CD

The CD accompanying this report contains data and analysis tools that can be used to assist with display and use of the results presented in the report, and to undertake further bioregionalisation analysis.

ArcGIS_files

The ArcGIS_files folder contains shapefiles with corresponding grids (raster datasets) and layer files for use in ArcGIS (created using ArcGIS version 9). These files can be used to display the results presented in the report, and to undertake further spatial analysis using GIS analysis tools. Table A2 gives details of the files provided. Further descriptions and metadata records are provided in Appendix III.

Use of the data in the ArcGIS_files directory is governed by the following conditions: 1. The data is provided for non-commercial use only. 2. Any publication derived using the datasets should acknowledge the Australian Antarctic Data Centre as having provided the data and the original source (see the relevant metadata record listed in Appendix III for the proper citation). Ben Raymond, Australian Government Antarctic Division October, 2006 [email protected]

TABLE A2: ArcGIS files provided on the CD Data/Feature Shapefile Grids Layer file (raster datasets) Chlorophyll a chlorophyll.shp chlorophyll chlorophyll.lyr Sea Ice proportion 15% ice_prop15.shp ice_prop15 ice_prop15.lyr Nitrate nox_mean.shp nox_mean nox_mean.lyr Silicate si_mean.shp si_mean si_mean.lyr Sea surface temperature sst.shp sst sst.lyr Southern ocean fronts sthn_ocean_fronts.shp - - Bathymetry bathy_areas.shp - bathy_areas_down_to _5000.lyr

bathy_areas_below_5 000.lyr Primary regionalisation regions_1_14.shp regions_1_14 regions_1_14.lyr Primary regionalisation scaled reunc_1_14_sc.shp reunc_1_14_sc reunc_1_14_sc.lyr uncertainty Secondary regionalisation ice reg_2_ice_12.shp reg_2_ice_12 reg_2_ice_12.lyr 12 clusters Secondary regionalisation reg_2_chl_15.shp reg_2_chl_15 reg_2_chl_15.lyr chlorophyll 15 clusters Secondary regionalisation ice reg2icechl_40.shp reg2icechl_40 reg2icechl_40.lyr and chlorophyll 40 clusters

xxi Matlab_code

The Matlab_code folder contains the code and data used to implement the bioregionalisation process outlined in the report.

Use of the data in the Matlab_code directory is governed by the following conditions: 1. The data is provided for non-commercial use only. 2. Any publication derived using the datasets should acknowledge the Australian Antarctic Data Centre as having provided the data and the original source (see the relevant metadata record listed in Appendix III for the proper citation). Ben Raymond, Australian Government Antarctic Division October, 2006 [email protected]

Installation: Copy the Matlab_code directory and its subdirectories to a suitable working location. You will need to have a local copy of R installed (see http://www.r-project.org), and also the R-(D)COM interface from http://lib.stat.cmu.edu/R/CRAN/other-software.html or other CRAN mirror site. R is used here to do the CLARA non-hierarchical clustering; note that this arrangement of using the Matlab R-link toolbox and R-DCOM server will probably only work on Windows platforms.

Usage: Run the do_bioreg script from within Matlab by typing 'do_bioreg' (see regions_1_14.png and regions_1_14.txt for example outputs). See comments within the script. A summary of the statistical properties of the resulting regions can be generated using cluster_properties.m (see region_properties_1_14.xls for example output). Different numbers of clusters can be obtained (from within Matlab): > n_groups=14; %change this 14 to the desired number (<40) > do_figure_group_colours % this extracts the groups from the dendrogram and plots the results An assessment of uncertainty (see report Sections 2.4 and 3.2) can be generated using do_uncert.m (see regions_uncertainty_1_14_scaled_nearest.png and regions_uncertainty_1_14_unscaled.png for example outputs). data/ -- this directory holds data files, in .mat (Matlab) format. CSV-text format copies are also included (zipped). Please see the readme.txt file in the data directory for dataset descriptions and conditions of use. code/ -- this directory holds some additional Matlab code that is required. Note that 3 external Matlab toolboxes are included: the m_map toolbox (http://www.eos.ubc.ca/~rich/map.html) written by Rich Pawlowicz ([email protected]), the Matlab R-link toolbox (http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=5051&objectType=FILE) written by Robert Henson ([email protected]), and the toolbox (http://www.biol.ttu.edu/Strauss/Matlab/matlab.htm) written by Richard Strauss ([email protected]).

xxii